In this paper we present our initial results in articulatory-toacoustic conversion based on tongue movement recordings using Deep Neural Networks (DNNs). Despite the fact that deep learning has revolutionized several fields, so far only a few researchers have applied DNNs for this task. Here, we compare various possible feature representation approaches combined with DNN-based regression. As the input, we recorded synchronized 2D ultrasound images and speech signals. The task of the DNN was to estimate Mel-Generalized Cepstrum-based Line Spectral Pair (MGC-LSP) coefficients, which then served as input to a standard pulse-noise vocoder for speech synthesis. As the raw ultrasound images have a relatively high resolution, we experimented with various feature selection and transformation approaches to reduce the size of the feature vectors. The synthetic speech signals resulting from the various DNN configurations were evaluated both using objective measures and a subjective listening test. We found that the representation that used several neighboring image frames in combination with a feature selection method was preferred both by the subjects taking part in the listening experiments, and in terms of the Normalized Mean Squared Error. Our results may be useful for creating Silent Speech Interface applications in the future.

No abstract

For articulatory-to-acoustic mapping using deep neural networks, typically spectral and excitation parameters of vocoders have been used as the training targets. However, vocoding often results in buzzy and muffled final speech quality. Therefore, in this paper on ultrasound-based articulatory-to-acoustic conversion, we use a flow-based neural vocoder (WaveGlow) pre-trained on a large amount of English and Hungarian speech data. The inputs of the convolutional neural network are ultrasound tongue images. The training target is the 80-dimensional mel-spectrogram, which results in a finer detailed spectral representation than the previously used 25-dimensional Mel-Generalized Cepstrum. From the output of the ultrasoundto-mel-spectrogram prediction, WaveGlow inference results in synthesized speech. We compare the proposed WaveGlowbased system with a continuous vocoder which does not use strict voiced/unvoiced decision when predicting F0. The results demonstrate that during the articulatory-to-acoustic mapping experiments, the WaveGlow neural vocoder produces significantly more natural synthesized speech than the baseline system. Besides, the advantage of WaveGlow is that F0 is included in the mel-spectrogram representation, and it is not necessary to predict the excitation separately.

Silent Speech Interface systems apply two different strategies to solve the articulatory-to-acoustic conversion task. The recognition-and-synthesis approach applies speech recognition techniques to map the articulatory data to a textual transcript, which is then converted to speech by a conventional text-tospeech system. The direct synthesis approach seeks to convert the articulatory information directly to speech synthesis (vocoder) parameters. In both cases, deep neural networks are an evident and popular choice to learn the mapping task. Recognizing that the learning of speech recognition and speech synthesis targets (acoustic model states vs. vocoder parameters) are two closely related tasks over the same ultrasound tongue image input, here we experiment with the multi-task training of deep neural networks, which seeks to solve the two tasks simultaneously. Our results show that the parallel learning of the two types of targets is indeed beneficial for both tasks. Moreover, we obtained further improvements by using multi-task training as a weight initialization step before task-specific training. Overall, we report a relative error rate reduction of about 7% in both the speech recognition and the speech synthesis tasks.

Recently it was shown that within the Silent Speech Interface (SSI) field, the prediction of F0 is possible from Ultrasound Tongue Images (UTI) as the articulatory input, using Deep Neural Networks for articulatory-to-acoustic mapping. Moreover, text-to-speech synthesizers were shown to produce higher quality speech when using a continuous pitch estimate, which takes non-zero pitch values even when voicing is not present. Therefore, in this paper on UTI-based SSI, we use a simple continuous F0 tracker which does not apply a strict voiced / unvoiced decision. Continuous vocoder parameters (ContF0, Maximum Voiced Frequency and Mel-Generalized Cepstrum) are predicted using a convolutional neural network, with UTI as input. The results demonstrate that during the articulatory-toacoustic mapping experiments, the continuous F0 is predicted with lower error, and the continuous vocoder produces slightly more natural synthesized speech than the baseline vocoder using standard discontinuous F0.