Olov Engwall

Combining MRI, EMA and EPG measurements in a three-dimensional tongue model ☆

A three-dimensional (3D) tongue model has been developed using MR images of a reference subject producing 44 artificially sustained Swedish articulations. Based on the difference in tongue shape between the articulations and a reference, the six linear parameters jaw height, tongue body, tongue dorsum, tongue tip, tongue advance and tongue width were determined using an ordered linear factor analysis controlled by articulatory measures. The first five factors explained 88% of the tongue data variance in the midsagittal plane and 78% in the 3D analysis. The six-parameter model is able to reconstruct the modelled articulations with an overall mean reconstruction error of 0.13 cm, and it specifically handles lateral differences and asymmetries in tongue shape. In order to correct articulations that were hyperarticulated due to the artificial sustaining in the magnetic resonance imaging (MRI) acquisition, the parameter values in the tongue model were readjusted based on a comparison of virtual and natural linguopalatal contact patterns, collected with electropalatography (EPG). Electromagnetic articulography (EMA) data was collected to control the kinematics of the tongue model for vowel-fricative sequences and an algorithm to handle surface contacts has been implemented, preventing the tongue from protruding through the palate and teeth.

Pronunciation feedback from real and virtual language teachers

The aim of this paper is to summarise how pronunciation feedback on the phoneme level should be given in computer-assisted pronunciation training (CAPT) in order to be effective. The study contains a literature survey of feedback in the language classroom, interviews with language teachers and their students about their attitudes towards pronunciation feedback, and observations of how feedback is given in their classrooms. The study was carried out using focus group meetings, individual semi-structured interviews and classroom observations. The feedback strategies that were advocated and observed in the study on pronunciation feedback from human teachers were implemented in a computer-animated language tutor giving articulation feedback. The virtual tutor was subsequently tested in a user trial and evaluated with a questionnaire. The article proposes several feedback strategies that would improve the pedagogical soundness of CAPT systems.

Wizard-of-Oz test of ARTUR: a computer-based speech training system with articulation correction

This study has been performed in order to test the human-machine interface of a computer-based speech training aid named ARTUR with the main feature that it can give suggestions on how to improve articulation. Two user groups were involved: three children aged 9-14 with extensive experience of speech training, and three children aged 6. All children had general language disorders.The study indicates that the present interface is usable without prior training or instructions, even for the younger children, although it needs some improvement to fit illiterate children. The granularity of the mesh that classifies mispronunciations was satisfactory, but can be developed further.

Designing the user interface of the computer-based speech training system ARTUR based on early user tests

This study has been performed in order to evaluate a prototype for the human – computer interface of a computer-based speech training aid named ARTUR. The main feature of the aid is that it can give suggestions on how to improve articulations. Two user groups were involved: three children aged 9 – 14 with extensive experience of speech training with therapists and computers, and three children aged 6, with little or no prior experience of computer-based speech training. All children had general language disorders. The study indicates that the present interface is usable without prior training or instructions, even for the younger children, but that more motivational factors should be introduced. The granularity of the mesh that classifies mispronunciations was satisfactory, but the flexibility and level of detail of the feedback should be developed further.

Analysis of and feedback on phonetic features in pronunciation training with a virtual teacher

Pronunciation errors may be caused by several different deviations from the target, such as voicing, intonation, insertions or deletions of segments, or that the articulators are placed incorrectly. Computer-animated pronunciation teachers could potentially provide important assistance on correcting all these types of deviations, but they have an additional benefit for articulatory errors. By making parts of the face transparent, they can show the correct position and shape of the tongue and provide audiovisual feedback on how to change erroneous articulations. Such a scenario however requires firstly that the learners current articulation can be estimated with precision and secondly that the learner is able to imitate the articulatory changes suggested in the audiovisual feedback. This article discusses both these aspects, with one experiment on estimating the important articulatory features from a speaker through acoustic-to-articulatory inversion and one user test with a virtual pronunciation teacher, in which the articulatory changes made by seven learners who receive audiovisual feedback are monitored using ultrasound imaging.

Audiovisual-to-articulatory inversion

It has been shown that acoustic-to-articulatory inversion, i.e. estimation of the articulatory configuration from the corresponding acoustic signal, can be greatly improved by adding visual features extracted from the speakers face. In order to make the inversion method usable in a realistic application, these features should be possible to obtain from a monocular frontal face video, where the speaker is not required to wear any special markers. In this study, we investigate the importance of visual cues for inversion. Experiments with motion capture data of the face show that important articulatory information can be extracted using only a few face measures that mimic the information that could be gained from a video-based method. We also show that the depth cue for these measures is not critical, which means that the relevant information can be extracted from a frontal video. A real video-based face feature extraction method is further presented, leading to similar improvements in inversion quality. Rather than tracking points on the face, it represents the appearance of the mouth area using independent component images. These findings are important for applications that need a simple audiovisual-to-articulatory inversion technique, e.g. articulatory phonetics training for second language learners or hearing-impaired persons.

Feedback management in the pronunciation training system ARTUR

This extended abstract discusses the development of a computer-assisted pronunciation training system that gives articulatory feedback, and in particular the management of feedback given to the user.

Mapping between acoustic and articulatory gestures

This paper proposes a definition for articulatory as well as acoustic gestures along with a method to segment the measured articulatory trajectories and acoustic waveforms into gestures. Using a simultaneously recorded acoustic-articulatory database, the gestures are detected based on finding critical points in the utterance, both in the acoustic and articulatory representations. The acoustic gestures are parameterized using 2-D cepstral coefficients. The articulatory trajectories are essentially the horizontal and vertical movements of Electromagnetic Articulography (EMA) coils placed on the tongue, jaw and lips along the midsagittal plane. The articulatory movements are parameterized using 2D-DCT using the same transformation that is applied on the acoustics. The relationship between the detected acoustic and articulatory gestures in terms of the timing as well as the shape is studied. In order to study this relationship further, acoustic-to-articulatory inversion is performed using GMM-based regression. The accuracy of predicting the articulatory trajectories from the acoustic waveforms are at par with state-of-the-art frame-based methods with dynamical constraints (with an average error of 1.45-1.55mm for the two speakers in the database). In order to evaluate the acoustic-to-articulatory inversion in a more intuitive manner, a method based on the error in estimated critical points is suggested. Using this method, it was noted that the estimated articulatory trajectories using the acoustic-to-articulatory inversion methods were still not accurate enough to be within the perceptual tolerance of audio-visual asynchrony.

Perceptual differentiation modeling explains phoneme mispronunciation by non-native speakers

One of the difficulties in second language (L2) learning is the weakness in discriminating between acoustic diversity within an L2 phoneme category and between different categories. In this paper, we describe a general method to quantitatively measure the perceptual difference between a group of native and individual nonnative speakers. Normally, this task includes subjective listening tests and/or a thorough linguistic study. We instead use a totally automated method based on a psycho-acoustic auditory model. For a certain phoneme class, we measure the similarity of the Euclidean space spanned by the power spectrum of a native speech signal and the Euclidean space spanned by the auditory model output. We do the same for a non-native speech signal. Comparing the two similarity measurements, we find problematic phonemes for a given speaker. To validate our method, we apply it to different groups of non-native speakers of various first language (L1) backgrounds. Our results are verified by the theoretical findings in literature obtained from linguistic studies.

Influence of vocal tract geometry simplifications on the numerical simulation of vowel sounds

For many years, the vocal tract shape has been approximated by one-dimensional (1D) area functions to study the production of voice. More recently, 3D approaches allow one to deal with the complex 3D vocal tract, although area-based 3D geometries of circular cross-section are still in use. However, little is known about the influence of performing such a simplification, and some alternatives may exist between these two extreme options. To this aim, several vocal tract geometry simplifications for vowels [ɑ], [i], and [u] are investigated in this work. Six cases are considered, consisting of realistic, elliptical, and circular cross-sections interpolated through a bent or straight midline. For frequencies below 4-5 kHz, the influence of bending and cross-sectional shape has been found weak, while above these values simplified bent vocal tracts with realistic cross-sections are necessary to correctly emulate higher-order mode propagation. To perform this study, the finite element method (FEM) has been used. FEM results have also been compared to a 3D multimodal method and to a classical 1D frequency domain model.