Publications by Korin Richmond

Forced alignment for speech synthesis traditionally aligns a phoneme sequence predetermined by the front-end text processing system. This sequence is not altered during alignment, i.e., it is forced, despite possibly being faulty. The consistency assumption is the assumption that these mistakes do not degrade models, as long as the mistakes are consistent across training and synthesis. We present evidence that in the alignment of both standard read prompts and spontaneous speech this phoneme sequence is often wrong, and that this is likely to have a negative impact on acoustic models. A lattice-based forced alignment system allowing for pronunciation variation is implemented, resulting in improved phoneme identity accuracy for both types of speech. A perceptual evaluation of HMM-based voices showed that spontaneous models trained on this improved alignment also improved standard synthesis, despite breaking the consistency assumption.

Keywords: speech synthesis, TTS, forced alignment, HMM

Although frequency analysis often leads us to a speech signal in the complex domain, the acoustic models we frequently use are designed for real-valued data. Phase is usually ignored or modelled separately from spectral amplitude. Here, we propose a complex-valued neural network (CVNN) for directly modelling the results of the frequency analysis in the complex domain (such as the complex amplitude). We also introduce a phase encoding technique to map real-valued data (e.g. cepstra or log amplitudes) into the complex domain so we can use the same CVNN processing seamlessly. In this paper, a fully complex-valued neural network, namely a neural network where all of the weight matrices, activation functions and learning algorithms are in the complex domain, is applied for speech synthesis. Results show its ability to model both complex-valued and real-valued data.

Keywords: complex-valued neural network, speech synthesis, complex amplitude, phase modelling

Join cost calculation has so far dealt exclusively with acoustic speech parameters, and a large number of distance metrics have previously been tested in conjunction with a wide variety of acoustic parameterisations. In contrast, we propose here to calculate distance in articulatory space. The motivation for this is simple: physical constraints mean a human talker's mouth cannot “jump” from one configuration to a different one, so smooth evolution of articulator positions would also seem desirable for a good candidate unit sequence. To test this, we built Festival Multisyn voices using a large articulatory-acoustic dataset. We first synthesised 460 TIMIT sentences and confirmed our articulatory join cost gives appreciably different unit sequences compared to the standard Multisyn acoustic join cost. A listening test (3 sets of 25 sentence pairs, 30 listeners) then showed our articulatory cost is preferred at a rate of 58% compared to the standard Multisyn acoustic join cost.

Keywords: speech synthesis, unit selection, electromagnetic articulography, join cost

It has recently been shown that deep neural networks (DNN) can improve the quality of statistical parametric speech synthesis (SPSS) when using a source-filter vocoder. Our own previous work has furthermore shown that a dynamic sinusoidal model (DSM) is also highly suited to DNN-based SPSS, whereby sinusoids may either be used themselves as a “direct parameterisation” (DIR), or they may be encoded using an “intermediate spectral parameterisation” (INT). The approach in that work was effectively to replace a decision tree with a neural network. However, waveform parameterisation and synthesis steps that have been developed to suit HMMs may not fully exploit DNN capabilities. Here, in contrast, we investigate ways to combine INT and DIR at the levels of both DNN modelling and waveform generation. For DNN training, we propose to use multi-task learning to model cepstra (from INT) and log amplitudes (from DIR) as primary and secondary tasks. Our results show combining these improves modelling accuracy for both tasks. Next, during synthesis, instead of discarding parameters from the second task, a fusion method using harmonic amplitudes derived from both tasks is applied. Preference tests show the proposed method gives improved performance, and that this applies to synthesising both with and without global variance parameters.

We describe a minimally-supervised method for computing a statistical shape space model of the palate surface. The model is created from a corpus of volumetric magnetic resonance imaging (MRI) scans collected from 12 speakers. We extract a 3D mesh of the palate from each speaker, then train the model using principal component analysis (PCA). The palate model is then tested using 3D MRI from another corpus and evaluated using a high-resolution optical scan. We find that the error is low even when only a handful of measured coordinates are available. In both cases, our approach yields promising results. It can be applied to extract the palate shape from MRI data, and could be useful to other analysis modalities, such as electromagnetic articulography (EMA) and ultrasound tongue imaging (UTI).

Keywords: vocal tract MRI, principal component analysis, palate model

Sinusoidal vocoders can generate high quality speech, but they have not been extensively applied to statistical parametric speech synthesis. This paper presents two ways for using dynamic sinusoidal models for statistical speech synthesis, enabling the sinusoid parameters to be modelled in HMMbased synthesis. In the first method, features extracted from a fixed- and low-dimensional, perception-based dynamic sinusoidal model (PDM) are statistically modelled directly. In the second method, we convert both static amplitude and dynamic slope from all the harmonics of a signal, which we term the Harmonic Dynamic Model (HDM), to intermediate parameters (regularised cepstral coefficients) for modelling. During synthesis, HDM is then used to reconstruct speech. We have compared the voice quality of these two methods to the STRAIGHT cepstrum-based vocoder with mixed excitation in formal listening tests. Our results show that HDM with intermediate parameters can generate comparable quality as STRAIGHT, while PDM direct modelling seems promising in terms of producing good speech quality without resorting to intermediate parameters such as cepstra.

This paper applies a dynamic sinusoidal synthesis model to statistical parametric speech synthesis (HTS). For this, we utilise regularised cepstral coefficients to represent both the static amplitude and dynamic slope of selected sinusoids for statistical modelling. During synthesis, a dynamic sinusoidal model is used to reconstruct speech. A preference test is conducted to compare the selection of different sinusoids for cepstral representation. Our results show that when integrated with HTS, a relatively small number of sinusoids selected according to a perceptual criterion can produce quality comparable to using all harmonics. A Mean Opinion Score (MOS) test shows that our proposed statistical system is preferred to one using mel-cepstra from pitch synchronous spectral analysis.

This paper presents a fixed- and low-dimensional, perceptually based dynamic sinusoidal model of speech referred to as PDM (Perceptual Dynamic Model). To decrease and fix the number of sinusoidal components typically used in the standard sinusoidal model, we propose to use only one dynamic sinusoidal component per critical band. For each band, the sinusoid with the maximum spectral amplitude is selected and associated with the centre frequency of that critical band. The model is expanded at low frequencies by incorporating sinusoids at the boundaries of the corresponding bands while at the higher frequencies a modulated noise component is used. A listening test is conducted to compare speech reconstructed with PDM and state-of-the-art models of speech, where all models are constrained to use an equal number of parameters. The results show that PDM is clearly preferred in terms of quality over the other systems.

This paper proposes an analysis method to separate the glottal source and vocal tract components of speech that is called Glottal Spectral Separation (GSS). This method can produce high-quality synthetic speech using an acoustic glottal source model. In the source-filter models commonly used in speech technology applications it is assumed the source is a spectrally flat excitation signal and the vocal tract filter can be represented by the spectral envelope of speech. Although this model can produce high-quality speech, it has limitations for voice transformation because it does not allow control over glottal parameters which are correlated with voice quality. The main problem with using a speech model that better represents the glottal source and the vocal tract filter is that current analysis methods for separating these components are not robust enough to produce the same speech quality as using a model based on the spectral envelope of speech. The proposed GSS method is an attempt to overcome this problem, and consists of the following three steps. Initially, the glottal source signal is estimated from the speech signal. Then, the speech spectrum is divided by the spectral envelope of the glottal source signal in order to remove the glottal source effects from the speech signal. Finally, the vocal tract transfer function is obtained by computing the spectral envelope of the resulting signal. In this work, the glottal source signal is represented using the Liljencrants-Fant model (LF-model). The experiments we present here show that the analysis-synthesis technique based on GSS can produce speech comparable to that of a high-quality vocoder that is based on the spectral envelope representation. However, it also permit control over voice qualities, namely to transform a modal voice into breathy and tense, by modifying the glottal parameters.

Keywords: Analytical models;Computational modeling;Estimation;Hidden Markov models;Mathematical model;Speech;Speech synthesis;Glottal spectral separation;LF-model;parametric speech synthesis;voice quality transformation

This paper presents an experimental comparison of a broad range of the leading vocoder types which have been previously described. We use a reference implementation of each of these to create stimuli for a listening test using copy synthesis. The listening test is performed using both Lombard and normal read speech stimuli, and with two types of question for comparison. Multi-dimensional Scaling (MDS) is conducted on the listener responses to analyse similarities in terms of quality between the vocoders. Our MDS and clustering results show that the vocoders which use a sinusoidal synthesis approach are perceptually distinguishable from the source-filter vocoders. To help further interpret the axes of the resulting MDS space, we test for correlations with standard acoustic quality metrics and find one axis is strongly correlated with PESQ scores. We also find both speech style and the format of the listening test question may influence test results. Finally, we also present preference test results which compare each vocoder with the natural speech.

The two measures typically used to assess the performance of an inversion mapping method, where the aim is to estimate what articulator movements gave rise to a given acoustic signal, are root mean squared (RMS) error and correlation. In this paper, we investigate whether “task-based” evaluation using an articulatory-controllable HMM-based speech synthesis system can give useful additional information to complement these measures. To assess the usefulness of this evaluation approach, we use articulator trajectories estimated by a range of different inversion mapping methods as input to the synthesiser, and measure their performance in the acoustic domain in terms of RMS error of the generated acoustic parameters and with a listening test involving 30 participants. We then compare these results with the standard RMS error and correlation measures calculated in the articulatory domain. Interestingly, in the acoustic evaluation we observe one method performs with no statistically significant difference from measured articulatory data, and cases where statistically significant differences between methods exist which are not reflected in the results of the two standard measures. From our results, we conclude such task-based evaluation can indeed provide interesting extra information, and gives a useful way to compare inversion methods.

Keywords: Inversion mapping, evaluation, HMM synthesis

The DoubleTalk articulatory corpus was collected at the Edinburgh Speech Production Facility (ESPF) using two synchronized Carstens AG500 electromagnetic articulometers. The first release of the corpus comprises orthographic transcriptions aligned at phrasal level to EMA and audio data for each of 6 mixed-dialect speaker pairs. It is available from the ESPF online archive. A variety of tasks were used to elicit a wide range of speech styles, including monologue (a modified Comma Gets a Cure and spontaneous story-telling), structured spontaneous dialogue (Map Task and Diapix), a wordlist task, a memory-recall task, and a shadowing task. In this session we will demo the corpus with various examples.

Keywords: discourse, EMA, spontaneous speech

In this paper, we present the recent progress in the MAGE project. MAGE is a library for realtime and interactive (reactive) parametric speech synthesis using hidden Markov models (HMMs). Here, it is broadened in order to support not only the standard acoustic features (spectrum and f0) to model and synthesize speech but also to combine acoustic and articulatory features, such as tongue, lips and jaw positions. Such an integration enables the user to have a straight forward and meaningful control space to intuitively modify the synthesized phones in real time only by configuring the position of the articulators.

Keywords: speech synthesis, reactive, articulators

In previous work we proposed a method to control the characteristics of synthetic speech flexibly by integrating articulatory features into a hidden Markov model (HMM) based parametric speech synthesiser. In this method, a unified acoustic-articulatory model is trained, and context-dependent linear transforms are used to model the dependency between the two feature streams. In this paper, we go significantly further and propose a feature-space-switched multiple regression HMM to improve the performance of articulatory control. A multiple regression HMM (MRHMM) is adopted to model the distribution of acoustic features, with articulatory features used as exogenous explanatory variables. A separate Gaussian mixture model (GMM) is introduced to model the articulatory space, and articulatory-to-acoustic regression matrices are trained for each component of this GMM, instead of for the context-dependent states in the HMM. Furthermore, we propose a task-specific context feature tailoring method to ensure compatibility between state context features and articulatory features that are manipulated at synthesis time. The proposed method is evaluated on two tasks, using a speech database with acoustic waveforms and articulatory movements recorded in parallel by electromagnetic articulography (EMA). In a vowel identity modification task, the new method achieves better performance when reconstructing target vowels by varying articulatory inputs than our previous approach. A second vowel creation task shows our new method is highly effective at producing a new vowel from appropriate articulatory representations which, even though no acoustic samples for this vowel are present in the training data, is shown to sound highly natural.

Abstract We demonstrate the workability of an experimental facility that is geared towards the acquisition of articulatory data from a variety of speech styles common in language use, by means of two synchronized electromagnetic articulography (EMA) devices. This approach synthesizes the advantages of real dialogue settings for speech research with a detailed description of the physiological reality of speech production. We describe the facility's method for acquiring synchronized audio streams of two speakers and the system that enables communication among control room technicians, experimenters and participants. Further, we demonstrate the feasibility of the approach by evaluating problems inherent to this specific setup: The first problem is the accuracy of temporal synchronization of the two {EMA} machines, the second is the severity of electromagnetic interference between the two machines. Our results suggest that the synchronization method used yields an accuracy of approximately 1 ms. Electromagnetic interference was derived from the complex-valued signal amplitudes. This dependent variable was analyzed as a function of the recording status - i.e. on/off - of the interfering machine's transmitters. The intermachine distance was varied between 1 m and 8.5 m. Results suggest that a distance of approximately 6.5 m is appropriate to achieve data quality comparable to that of single speaker recordings.

Electromagnetic articulography (EMA) captures the position and orientation of a number of markers, attached to the articulators, during speech. As such, it performs the same function for speech that conventional motion capture does for full-body movements acquired with optical modalities, a long-time staple technique of the animation industry. In this paper, EMA data is processed from a motion-capture perspective and applied to the visualization of an existing multimodal corpus of articulatory data, creating a kinematic 3D model of the tongue and teeth by adapting a conventional motion capture based animation paradigm. This is accomplished using off-the-shelf, open-source software. Such an animated model can then be easily integrated into multimedia applications as a digital asset, allowing the analysis of speech production in an intuitive and accessible manner. The processing of the EMA data, its co-registration with 3D data from vocal tract magnetic resonance imaging (MRI) and dental scans, and the modeling workflow are presented in detail, and several issues discussed.

Keywords: speech production, articulatory data, electromagnetic articulography, vocal tract, motion capture, visualization

Speech sound disorders (SSD) are the most common communication impairment in childhood, and can hamper social development and learning. Current speech therapy interventions rely predominantly on the auditory skills of the child, as little technology is available to assist in diagnosis and therapy of SSDs. Realtime visualisation of tongue movements has the potential to bring enormous benefit to speech therapy. Ultrasound scanning offers this possibility, although its display may be hard to interpret. Our ultimate goal is to exploit ultrasound to track tongue movement, while displaying a simplified, diagrammatic vocal tract that is easier for the user to interpret. In this paper, we outline a general approach to this problem, combining a latent space model with a dimensionality reducing model of vocal tract shapes. We assess the feasibility of this approach using magnetic resonance imaging (MRI) scans to train a model of vocal tract shapes, which is animated using electromagnetic articulography (EMA) data from the same speaker.

Keywords: Ultrasound, speech therapy, vocal tract visualisation

This paper presents a method to produce a new vowel by articulatory control in hidden Markov model (HMM) based parametric speech synthesis. A multiple regression HMM (MRHMM) is adopted to model the distribution of acoustic features, with articulatory features used as external auxiliary variables. The dependency between acoustic and articulatory features is modelled by a group of linear transforms that are either estimated context-dependently or determined by the distribution of articulatory features. Vowel identity is removed from the set of context features used to ensure compatibility between the context-dependent model parameters and the articulatory features of a new vowel. At synthesis time, acoustic features are predicted according to the input articulatory features as well as context information. With an appropriate articulatory feature sequence, a new vowel can be generated even when it does not exist in the training set. Experimental results show this method is effective in creating the English vowel /2/ by articulatory control without using any acoustic samples of this vowel.

Keywords: Speech synthesis, articulatory features, multiple-regression hidden Markov model

We implement two deep architectures for the acoustic-articulatory inversion mapping problem: a deep neural network and a deep trajectory mixture density network. We find that in both cases, deep architectures produce more accurate predictions than shallow architectures and that this is due to the higher expressive capability of a deep model and not a consequence of adding more adjustable parameters. We also find that a deep trajectory mixture density network is able to obtain better inversion accuracies than smoothing the results of a deep neural network. Our best model obtained an average root mean square error of 0.885 mm on the MNGU0 test dataset.

Keywords: Articulatory inversion, deep neural network, deep belief network, deep regression network, pretraining

This paper announces the availability of the magnetic resonance imaging (MRI) subset of the mngu0 corpus, a collection of articulatory speech data from one speaker containing different modalities. This subset comprises volumetric MRI scans of the speaker's vocal tract during sustained production of vowels and consonants, as well as dynamic mid-sagittal scans of repetitive consonant-vowel (CV) syllable production. For reference, high-quality acoustic recordings of the speech material are also available. The raw data are made freely available for research purposes.

Keywords: audio recording; magnetic resonance imaging; speech processing

In this work, we implement a deep belief network for the acoustic-articulatory inversion mapping problem. We find that adding up to 3 hidden-layers improves inversion accuracy. We also show that this improvement is due to the higher ex- pressive capability of a deep model and not a consequence of adding more adjustable parameters. Additionally, we show unsupervised pretraining of the sys- tem improves its performance in all cases, even for a 1 hidden-layer model. Our implementation obtained an average root mean square error of 0.95 mm on the MNGU0 test dataset, beating all previously published results.

This paper serves as an initial announcement of the availability of a corpus of articulatory data called mngu0. This corpus will ultimately consist of a collection of multiple sources of articulatory data acquired from a single speaker: electromagnetic articulography (EMA), audio, video, volumetric MRI scans, and 3D scans of dental impressions. This data will be provided free for research use. In this first stage of the release, we are making available one subset of EMA data, consisting of more than 1,300 phonetically diverse utterances recorded with a Carstens AG500 electromagnetic articulograph. Distribution of mngu0 will be managed by a dedicated “forum-style” web site. This paper both outlines the general goals motivating the distribution of the data and the creation of the mngu0 web forum, and also provides a description of the EMA data contained in this initial release.

This paper proposes a novel framework that enables us to manipulate and control formants in HMM-based speech synthesis. In this framework, the dependency between formants and spectral features is modelled by piecewise linear transforms; formant parameters are effectively mapped by these to the means of Gaussian distributions over the spectral synthesis parameters. The spectral envelope features generated under the influence of formants in this way may then be passed to high-quality vocoders to generate the speech waveform. This provides two major advantages over conventional frameworks. First, we can achieve spectral modification by changing formants only in those parts where we want control, whereas the user must specify all formants manually in conventional formant synthesisers (e.g. Klatt). Second, this can produce high-quality speech. Our results show the proposed method can control vowels in the synthesized speech by manipulating F 1 and F 2 without any degradation in synthesis quality.

In previous work, we have proposed a method to control the characteristics of synthetic speech flexibly by integrating articulatory features into hidden Markov model (HMM) based parametric speech synthesis. A unified acoustic-articulatory model was trained and a piecewise linear transform was adopted to describe the dependency between these two feature streams. The transform matrices were trained for each HMM state and were tied based on each state's context. In this paper, an improved acoustic-articulatory modelling method is proposed. A Gaussian mixture model (GMM) is introduced to model the articulatory space and the cross-stream transform matrices are trained for each Gaussian mixture instead of context-dependently. This means the dependency relationship can vary with the change of articulatory features flexibly. Our results show this method improves the effectiveness of control over vowel quality by modifing articulatory trajectories without degrading naturalness.

A major factor which causes a deterioration in speech quality in HMM-based speech synthesis is the use of a simple delta pulse signal to generate the excitation of voiced speech. This paper sets out a new approach to using an acoustic glottal source model in HMM-based synthesisers instead of the traditional pulse signal. The goal is to improve speech quality and to better model and transform voice characteristics. We have found the new method decreases buzziness and also improves prosodic modelling. A perceptual evaluation has supported this finding by showing a 55.6% preference for the new system, as against the baseline. This improvement, while not being as significant as we had initially expected, does encourage us to work on developing the proposed speech synthesiser further.

This paper presents an investigation into predicting the movement of a speaker's mouth from text input using hidden Markov models (HMM). A corpus of human articulatory movements, recorded by electromagnetic articulography (EMA), is used to train HMMs. To predict articulatory movements for input text, a suitable model sequence is selected and a maximum-likelihood parameter generation (MLPG) algorithm is used to generate output articulatory trajectories. Unified acoustic-articulatory HMMs are introduced to integrate acoustic features when an acoustic signal is also provided with the input text. Several aspects of this method are analyzed in this paper, including the effectiveness of context-dependent modeling, the role of supplementary acoustic input, and the appropriateness of certain model structures for the unified acoustic-articulatory models. When text is the sole input, we find that fully context-dependent models significantly outperform monophone and quinphone models, achieving an average root mean square (RMS) error of 1.945 mm and an average correlation coefficient of 0.600. When both text and acoustic features are given as input to the system, the difference between the performance of quinphone models and fully context-dependent models is no longer significant. The best performance overall is achieved using unified acoustic-articulatory quinphone HMMs with separate clustering of acoustic and articulatory model parameters, a synchronous-state sequence, and a dependent-feature model structure, with an RMS error of 0.900 mm and a correlation coefficient of 0.855 on average. Finally, we also apply the same quinphone HMMs to the acoustic-articulatory, or inversion, mapping problem, where only acoustic input is available. An average root mean square (RMS) error of 1.076 mm and an average correlation coefficient of 0.812 are achieved. Taken together, our results demonstrate how text and acoustic inputs both contribute to the prediction of articulatory movements in the method used.

Keywords: Hidden Markov model; Articulatory features; Parameter generation

In this paper we present a method to predict the movement of a speaker's mouth from text input using hidden Markov models (HMM). We have used a corpus of human articulatory movements, recorded by electromagnetic articulography (EMA), to train HMMs. To predict articulatory movements from text, a suitable model sequence is selected and the maximum-likelihood parameter generation (MLPG) algorithm is used to generate output articulatory trajectories. In our experiments, we find that fully context-dependent models outperform monophone and quinphone models, achieving an average root mean square (RMS) error of 1.945mm when state durations are predicted from text, and 0.872mm when natural state durations are used. Finally, we go on to analyze the prediction error for different EMA dimensions and phone types. We find a clear pattern emerges that the movements of so-called critical articulators can be predicted more accurately than the average performance.

Keywords: Hidden Markov model, articulatory features, parameter generation, critical articulators

We present a new phone-dependent feature weighting scheme that can be used to map articulatory configurations (e.g. EMA) onto vocal tract spectra (e.g. MFCC) through table lookup. The approach consists of assigning feature weights according to a feature's ability to predict the acoustic distance between frames. Since an articulator's predictive accuracy is phone-dependent (e.g., lip location is a better predictor for bilabial sounds than for palatal sounds), a unique weight vector is found for each phone. Inspection of the weights reveals a correspondence with the expected critical articulators for many phones. The proposed method reduces overall cepstral error by 6% when compared to a uniform weighting scheme. Vowels show the greatest benefit, though improvements occur for 80% of the tested phones.

Keywords: speech production, speech synthesis

Combilex is a high-quality lexicon that has been developed specifically for speech technology purposes and recently released by CSTR. Combilex benefits from many advanced features. This paper explores one of these: the ability to generate fully-specified transcriptions for morphologically derived words automatically. This functionality was originally implemented to encode the pronunciations of derived words in terms of their constituent morphemes, thus accelerating lexicon development and ensuring a high level of consistency. In this paper, we propose this method of modelling pronunciations can be exploited further by combining it with a morphological parser, thus yielding a method to generate full transcriptions for unknown derived words. Not only could this accelerate adding new derived words to Combilex, but it could also serve as an alternative to conventional letter-to-sound rules. This paper presents preliminary work indicating this is a promising direction.

Keywords: combilex lexicon, letter-to-sound rules, grapheme-to-phoneme conversion, morphological decomposition

Control over voice quality, e.g. breathy and tense voice, is important for speech synthesis applications. For example, transformations can be used to modify aspects of the voice re- lated to speaker's identity and to improve expressiveness. How- ever, it is hard to modify voice characteristics of the synthetic speech, without degrading speech quality. State-of-the-art sta- tistical speech synthesisers, in particular, do not typically al- low control over parameters of the glottal source, which are strongly correlated with voice quality. Consequently, the con- trol of voice characteristics in these systems is limited. In con- trast, the HMM-based speech synthesiser proposed in this paper uses an acoustic glottal source model. The system passes the glottal signal through a whitening filter to obtain the excitation of voiced sounds. This technique, called glottal post-filtering, allows to transform voice characteristics of the synthetic speech by modifying the source model parameters. We evaluated the proposed synthesiser in a perceptual ex- periment, in terms of speech naturalness, intelligibility, and similarity to the original speaker's voice. The results show that it performed as well as a HMM-based synthesiser, which generates the speech signal with a commonly used high-quality speech vocoder.

Keywords: HMM-based speech synthesis, voice quality, glottal post-filter

This paper presents a comparison between a hidden Markov model (HMM) based method and a novel artificial neural network (ANN) based method for lip synchronisation. Both model types were trained on motion tracking data, and a perceptual evaluation was carried out comparing the output of the models, both to each other and to the original tracked data. It was found that the ANN-based method was judged significantly better than the HMM based method. Furthermore, the original data was not judged significantly better than the output of the ANN method.

Keywords: hidden Markov model (HMM), mixture density network, lip synchronisation, inversion mapping

The EPSRC‐funded Edinburgh Speech Production is built around two synchronized Carstens AG500 electromagnetic articulographs (EMAs) in order to capture articulatory∕acoustic data from spontaneous dialogue. An initial articulatory corpus was designed with two aims. The first was to elicit a range of speech styles∕registers from speakers, and therefore provide an alternative to fully scripted corpora. The second was to extend the corpus beyond monologue, by using tasks that promote natural discourse and interaction. A subsidiary driver was to use dialects from outwith North America: dialogues paired up a Scottish English and a Southern British English speaker. Tasks. Monologue: Story reading of “Comma Gets a Cure” [Honorof et al. (2000)], lexical sets [Wells (1982)], spontaneous story telling, diadochokinetic tasks. Dialogue: Map tasks [Anderson et al. (1991)], “Spot the Difference” picture tasks [Bradlow et al. (2007)], story‐recall. Shadowing of the spontaneous story telling by the second participant. Each dialogue session includes approximately 30 min of speech, and there are acoustics‐only baseline materials. We will introduce the corpus and highlight the role of articulatory production data in helping provide a fuller understanding of various spontaneous speech phenomena by presenting examples of naturally occurring covert speech errors, accent accommodation, turn taking negotiation, and shadowing.

In this paper, we apply our inversion mapping method, the trajectory mixture density network (TMDN), to a new corpus of articulatory data, recorded with a Carstens AG500 electromagnetic articulograph. This new data set, mngu0, is relatively large and phonetically rich, among other beneficial characteristics. We obtain good results, with a root mean square (RMS) error of only 0.99mm. This compares very well with our previous lowest result of 1.54mm RMS error for equivalent coils of the MOCHA fsew0 EMA data. We interpret this as showing the mngu0 data set is potentially more consistent than the fsew0 data set, and is very useful for research which calls for articulatory trajectory data. It also supports our view that the TMDN is very much suited to the inversion mapping problem.

Keywords: acoustic-articulatory inversion mapping, neural network

Combilex is a high quality pronunciation lexicon aimed at speech technology applications that has recently been released by CSTR. Combilex benefits from several advanced features. This paper evaluates one of these: the explicit alignment of phones to graphemes in a word. This alignment can help to rapidly develop robust and accurate letter-to-sound (LTS) rules, without needing to rely on automatic alignment methods. To evaluate this, we used Festival's LTS module, comparing its standard automatic alignment with Combilex's explicit alignment. Our results show using Combilex's alignment improves LTS accuracy: 86.50% words correct as opposed to 84.49%, with our most general form of lexicon. In addition, building LTS models is greatly accelerated, as the need to list allowed alignments is removed. Finally, loose comparison with other studies indicates Combilex is a superior quality lexicon in terms of consistency and size.

Keywords: combilex, letter-to-sound rules, grapheme-to-phoneme conversion

As part of ongoing research towards integrating an articulatory synthesizer into a text-to-speech (TTS) framework, a corpus of German utterances recorded with electromagnetic articulography (EMA) is resynthesized to provide training data for statistical models. The resynthesis is based on a measure of similarity between the original and resynthesized EMA trajectories, weighted by articulatory relevance. Preliminary results are discussed and future work outlined.

Keywords: articulatory speech synthesis, copy synthesis, electromagnetic articulography, EMA

This paper presents an investigation of ways to integrate articulatory features into Hidden Markov Model (HMM)-based parametric speech synthesis, primarily with the aim of improving the performance of acoustic parameter generation. The joint distribution of acoustic and articulatory features is estimated during training and is then used for parameter generation at synthesis time in conjunction with a maximum-likelihood criterion. Different model structures are explored to allow the articulatory features to influence acoustic modeling: model clustering, state synchrony and cross-stream feature dependency. The results of objective evaluation show that the accuracy of acoustic parameter prediction can be improved when shared clustering and asynchronous-state model structures are adopted for combined acoustic and articulatory features. More significantly, our experiments demonstrate that modeling the dependency between these two feature streams can make speech synthesis more flexible. The characteristics of synthetic speech can be easily controlled by modifying generated articulatory features as part of the process of acoustic parameter generation.

A major cause of degradation of speech quality in HMM-based speech synthesis is the use of a simple delta pulse signal to generate the excitation of voiced speech. This paper describes a new approach to using an acoustic glottal source model in HMM-based synthesisers. The goal is to improve speech quality and parametric flexibility to better model and transform voice characteristics.

As part of ongoing work to integrate an articulatory synthesizer into a modular TTS platform, a method is presented which allows gestural timings to be generated automatically from EMA data. Further work is outlined which will adapt the vocal tract model and phoneset to English using new articulatory data, and use statistical trajectory models.

This paper presents a method to control the characteristics of synthetic speech flexibly by integrating articulatory features into a Hidden Markov Model (HMM)-based parametric speech synthesis system. In contrast to model adaptation and interpolation approaches for speaking style control, this method is driven by phonetic knowledge, and target speech samples are not required. The joint distribution of parallel acoustic and articulatory features considering cross-stream feature dependency is estimated. At synthesis time, acoustic and articulatory features are generated simultaneously based on the maximum-likelihood criterion. The synthetic speech can be controlled flexibly by modifying the generated articulatory features according to arbitrary phonetic rules in the parameter generation process. Our experiments show that the proposed method is effective in both changing the overall character of synthesized speech and in controlling the quality of a specific vowel.

We present a method for predicting the midsagittal tongue contour from the locations of a few landmarks (metal pellets) on the tongue surface, as used in articulatory databases such as MOCHA and the Wisconsin XRDB. Our method learns a mapping using ground-truth tongue contours derived from ultrasound data and drastically improves over spline interpolation. We also determine the optimal locations of the landmarks, and the number of landmarks required to achieve a desired prediction error: 3-4 landmarks are enough to achieve 0.3-0.2 mm error per point on the tongue.

This paper presents a method to control the characteristics of synthetic speech flexibly by integrating articulatory features into a Hidden Markov Model (HMM)-based parametric speech synthesis system. In contrast to model adaptation and interpolation approaches for speaking style control, this method is driven by phonetic knowledge, and target speech samples are not required. The joint distribution of parallel acoustic and articulatory features considering cross-stream feature dependency is estimated. At synthesis time, acoustic and articulatory features are generated simultaneously based on the maximum-likelihood criterion. The synthetic speech can be controlled flexibly by modifying the generated articulatory features according to arbitrary phonetic rules in the parameter generation process. Our experiments show that the proposed method is effective in both changing the overall character of synthesized speech and in controlling the quality of a specific vowel.

We have previously proposed a trajectory model which is based on a mixture density network (MDN) trained with target variables augmented with dynamic features together with an algorithm for estimating maximum likelihood trajectories which respects the constraints between those features. In this paper, we have extended that model to allow diagonal covariance matrices and multiple mixture components in the trajectory MDN output probability density functions. We have evaluated this extended model on an inversion mapping task and found the trajectory model works well, outperforming smoothing of equivalent trajectories using low-pass filtering. Increasing the number of mixture components in the TMDN improves results further.

This paper proposes the idea that by viewing an inversion mapping MLP from a Multitask Learning perspective, we may be able to relax two constraints which are inherent in using electromagnetic articulography as a source of articulatory information for speech technology purposes. As a first step to evaluating this idea, we perform an inversion mapping experiment in an attempt to ascertain whether the hidden layer of a “multitask” MLP can act beneficially as a hidden representation that is shared between inversion mapping subtasks for multiple articulatory targets. Our results in the case of the tongue dorsum x-coordinate indicate this is indeed the case and show good promise. Results for the tongue dorsum y-coordinate however are not so clear-cut, and will require further investigation.

This paper describes selected aspects of the Festival Multisyn entry to the Blizzard Challenge 2007. We provide an overview of the process of building the three required voices from the speech data provided. This paper focuses on new features of Multisyn which are currently under development and which have been employed in the system used for this Blizzard Challenge. These differences are the application of a more flexible phonetic lattice representation during forced alignment labelling and the use of a pitch accent target cost component. Finally, we also examine aspects of the speech data provided for this year's Blizzard Challenge and raise certain issues for discussion concerning the aim of comparing voices made with differing subsets of the data provided.

Although much is known about how speech is produced, and research into speech production has resulted in measured articulatory data, feature systems of different kinds and numerous models, speech production knowledge is almost totally ignored in current mainstream approaches to automatic speech recognition. Representations of speech production allow simple explanations for many phenomena observed in speech which cannot be easily analyzed from either acoustic signal or phonetic transcription alone. In this article, we provide a survey of a growing body of work in which such representations are used to improve automatic speech recognition.

This paper proposes the use of the Liljencrants-Fant model (LF-model) to represent the glottal source signal in HMM-based speech synthesis systems. These systems generally use a pulse train to model the periodicity of the excitation signal of voiced speech. However, this model produces a strong and uniform harmonic structure throughout the spectrum of the excitation which makes the synthetic speech sound buzzy. The use of a mixed band excitation and phase manipulation reduces this effect but it can result in degradation of the speech quality if the noise component is not weighted carefully. In turn, the LF-waveform has a decaying spectrum at higher frequencies, which is more similar to the real glottal source excitation signal. We conducted a perceptual experiment to test the hypothesis that the LF-model can perform as well as or better than the pulse train in a HMM-based speech synthesizer. In the synthesis, we used the mean values of the LF-parameters, calculated by measurements of the recorded speech. The result of this study is important not only regarding the improvement in speech quality of these type of systems, but also because the LF-model can be used to model many characteristics of the glottal source, such as voice quality, which are important for voice transformation and generation of expressive speech.

We present the implementation and evaluation of an open-domain unit selection speech synthesis engine designed to be flexible enough to encourage further unit selection research and allow rapid voice development by users with minimal speech synthesis knowledge and experience. We address the issues of automatically processing speech data into a usable voice using automatic segmentation techniques and how the knowledge obtained at labelling time can be exploited at synthesis time. We describe target cost and join cost implementation for such a system and describe the outcome of building voices with a number of different sized datasets. We show that, in a competitive evaluation, voices built using this technology compare favourably to other systems.

Current lexica for speech technology typically contain much redundancy, while omitting useful information. A comparison with lexica in other media and for other purposes is instructive, as it highlights some features we may borrow for text-to-speech and speech recognition lexica. We describe some aspects of the new lexicon we are producing, Combilex, whose structure and implementation is specifically designed to reduce redundancy and improve the representation of productive elements of English. Most importantly, many English words are predictable derivations of baseforms, or compounds. Storing the lexicon as a combination of baseforms and derivational rules speeds up lexicon development, and improves coverage and maintainability.

This paper describes the process of building unit selection voices for the Festival Multisyn engine using the ATR dataset provided for the Blizzard Challenge 2006. We begin by discussing recent improvements that we have made to the Multisyn voice building process, prompted by our participation in the Blizzard Challenge 2006. We then go on to discuss our interpretation of the results observed. Finally, we conclude with some comments and suggestions for the formulation of future Blizzard Challenges.

This paper proposes a trajectory model which is based on a mixture density network trained with target features augmented with dynamic features together with an algorithm for estimating maximum likelihood trajectories which respects constraints between the static and derived dynamic features. This model was evaluated on an inversion mapping task. We found the introduction of the trajectory model successfully reduced root mean square error by up to 7.5%, as well as increasing correlation scores.

This paper describes the process of building unit selection voices for the Festival Multisyn engine using four ARCTIC datasets, as part of the Blizzard evaluation challenge. The build process is almost entirely automatic, with very little need for human intervention. We discuss the difference in the evaluation results for each voice and evaluate the suitability of the ARCTIC datasets for building this type of voice.

The goal of this project was to build a unit selection voice that could portray emotions with varying intensities. A suitable definition of an emotion was developed along with a descriptive framework that supported the work carried out. A single speaker was recorded portraying happy and angry speaking styles. Additionally a neutral database was also recorded. A target cost function was implemented that chose units according to emotion mark-up in the database. The Dictionary of Affect supported the emotional target cost function by providing an emotion rating for words in the target utterance. If a word was particularly 'emotional', units from that emotion were favoured. In addition intensity could be varied which resulted in a bias to select a greater number emotional units. A perceptual evaluation was carried out and subjects were able to recognise reliably emotions with varying amounts of emotional units present in the target utterance.

Peña, Bonatti, Nespor and Mehler(2002) investigated an artificial language where the structure of words was determined by nonadjacent dependencies between syllables. They found that segmentation of continuous speech could proceed on the basis of these dependencies. However, Peña et al.'s artificial language contained a confound in terms of phonology, in that the dependent syllables began with plosives and the intervening syllables began with continuants. We consider three hypotheses concerning the role of phonology in speech segmentation in this task: (1) participants may recruit probabilistic phonotactic information from their native language to the artificial language learning task; (2) phonetic properties of the stimuli, such as the gaps that precede unvoiced plosives, can influence segmentation; and (3) grouping by phonological similarity between dependent syllables contributes to learning the dependency. In a series of experiments controlling the phonological and statistical structure of the language, we found that segmentation performance is influenced by the three factors in different degrees. Learning of non-adjacent dependencies did not occur when (3) is eliminated. We suggest that phonological processing provides a fundamental contribution to distributional analysis.

Conversational interfaces allow human users to use spoken language to interact with computer-based information services. In this paper, we examine the potential for personalizing speech-based human-computer interaction according to the user's gender and age. We describe a system that uses acoustic features of the user's speech to automatically estimate these physical characteristics. We discuss the difficulties of implementing this process in relation to the high level of environmental noise that is typical of mobile human-computer interaction.

This paper describes version 2 of the Festival speech synthesis system. Festival 2 provides a development environment for concatenative speech synthesis, and now includes a general purpose unit selection speech synthesis engine. We discuss various aspects of unit selection speech synthesis, focusing on the research issues that relate to voice design and the automation of the voice development process.

This paper presents an experimental comparison of the performance of the multilayer perceptron (MLP) with that of the mixture density network (MDN) for an acoustic-to-articulatory mapping task. A corpus of acoustic-articulatory data recorded by electromagnetic articulography (EMA) for a single speaker was used as training and test data for this purpose. In theory, the MDN is able to provide a richer, more flexible description of the target variables in response to a given input vector than the least-squares trained MLP. Our results show that the mean likelihoods of the target articulatory parameters for an unseen test set were indeed consistently higher with the MDN than with the MLP. The increase ranged from approximately 3% to 22%, depending on the articulatory channel in question. On the basis of these results, we argue that using a more flexible description of the target domain, such as that offered by the MDN, can prove beneficial when modelling the acoustic-to-articulatory mapping.

A successful method for inferring articulation from the acoustic speech signal would find many applications: low bit-rate speech coding, visual representation of speech, and the possibility of improved automatic speech recognition to name but a few. It is unsurprising, therefore, that researchers have been investigating the acoustic-to-articulatory inversion mapping for several decades now. A great variety of approaches and models have been applied to the problem. Unfortunately, the overwhelming majority of these attempts have faced difficulties in satisfactorily assessing performance in terms of genuine human articulation. However, technologies such as electromagnetic articulography (EMA) mean that measurement of human articulation during speech has become increasingly accessible. Crucially, a large corpus of acoustic-articulatory data during phonetically-diverse, continuous speech has recently been recorded at Queen Margaret College, Edinburgh. One of the primary motivations of this thesis is to exploit the availability of this remarkable resource. Among the data-driven models which have been employed in previous studies, the feedforward multilayer perceptron (MLP) in particular has been used several times with promising results. Researchers have cited advantages in terms of memory requirement and execution speed as a significant factor motivating their use. Furthermore, the MLP is well known as a universal function approximator; an MLP of suitable form can in theory represent any arbitrary mapping function. Therefore, using an MLP in conjunction with the relatively large quantities of acoustic-articulatory data arguably represents a promising and useful first research step for the current thesis, and a significant part of this thesis is occupied with doing this. Having demonstrated an MLP which performs well enough to provide a reasonable baseline, we go on to critically evaluate the suitability of the MLP for the inversion mapping. The aim is to find ways to improve modelling accuracy further. Considering what model of the target articulatory domain is provided in the MLP is key in this respect. It has been shown that the outputs of an MLP trained with the sum-of-squares error function approximate the mean of the target data points conditioned on the input vector. In many situations, this is an appropriate and sufficient solution. In other cases, however, this conditional mean is an inconveniently limiting model of data in the target domain, particularly for ill-posed problems where the mapping may be multi-valued. Substantial evidence exists which shows that multiple articulatory configurations are able to produce the same acoustic signal. This means that a system intended to map from a point in acoustic space can be faced with multiple candidate articulatory configurations. Therefore, despite the impressive ability of the MLP to model mapping functions, it may prove inadequate in certain respects for performing the acoustic-to-articulatory inversion mapping. Mixture density networks (MDN) provide a principled method to model arbitrary probability density functions over the target domain, conditioned on the input vector. In theory, therefore, the MDN offers a superior model of the target domain compared to the MLP. We hypothesise that this advantage will prove beneficial in the case of the acoustic-to-articulatory inversion mapping. Accordingly, this thesis aims to test this hypothesis and directly compare the performance of MDN with MLP on exactly the same acoustic-to-articulatory inversion task.

In this paper we show that there is measurable information in the articulatory system which can help to disambiguate the acoustic signal. We measure directly the movement of the lips, tongue, jaw, velum and larynx and parameterise this articulatory feature space using principle components analysis. The parameterisation is developed and evaluated using a speaker dependent phone recognition task on a specially recorded TIMIT corpus of 460 sentences. The results show that there is useful supplementary information contained in the articulatory data which yields a small but significant improvement in phone recognition accuracy of 2%. However, preliminary attempts to estimate the articulatory data from the acoustic signal and use this to supplement the acoustic input have not yielded any significant improvement in phone accuracy.

In this paper we describe a speech recognition system using linear dynamic models and articulatory features. Experiments are reported in which measured articulation from the MOCHA corpus has been used, along with those where the articulatory parameters are estimated from the speech signal using a recurrent neural network.

We describe recent work on two new automatic speech recognition systems. The first part of this paper describes the components of a system based on phonological features (which we call EspressoA) in which the values of these features are estimated from the speech signal before being used as the basis for recognition. In the second part of the paper, another system (which we call EspressoB) is described in which articulatory parameters are used instead of phonological features and a linear dynamical system model is used to perform recognition from automatically estimated values of these articulatory parameters.

This paper reports on present work, in which a recurrent neural network is trained to estimate `velum height' during continuous speech. Parallel acoustic-articulatory data comprising more than 400 read TIMIT sentences is obtained using electromagnetic articulography (EMA). This data is processed and used as training data for a range of neural network sizes. The network demonstrating the highest accuracy is identified. This performance is then evaluated in detail by analysing the network's output for each phonetic segment contained in 50 hand-labelled utterances set aside for testing purposes.

We describe here an algorithm for detecting subject boundaries within text based on a statistical lexical similarity measure. Hearst has already tackled this problem with good results (Hearst, 1994). One of her main assumptions is that a change in subject is accompanied by a change in vocabulary. Using this assumption, but by introducing a new measure of word significance, we have been able to build a robust and reliable algorithm which exhibits improved accuracy without sacrificing language independency.