Silent speech recognition (SSR) converts non-audio information such as articulatory movements into text. SSR has the potential to enable persons with laryngectomy to communicate through natural spoken expression. Current SSR systems have largely relied on speaker-dependent recognition models. The high degree of variability in articulatory patterns across different speakers has been a barrier for developing effective speaker-independent SSR approaches. Speaker-independent SSR approaches, however, are critical for reducing the amount of training data required from each speaker. In this paper, we investigate speaker-independent SSR from the movements of flesh points on tongue and lip with articulatory normalization methods that reduce the inter-speaker variation. To minimize the across-speaker physiological differences of the articulators, we propose Procrustes matching-based articulatory normalization by removing locational, rotational, and scaling differences. To further normalize the articulatory data, we apply feature-space maximum likelihood linear regression and i-vector. In this paper, we adopt a bidirectional long short term memory recurrent neural network (BLSTM) as an articulatory model to effectively model the articulatory movements with long-range articulatory history. A silent speech data set with flesh points was collected using an electromagnetic articulograph (EMA) from twelve healthy and two laryngectomized English speakers. Experimental results showed the effectiveness of our speaker-independent SSR approaches on healthy as well as laryngectomy speakers. In addition, BLSTM outperformed standard deep neural network. The best performance was obtained by BLSTM with all the three normalization approaches combined.

Vocal folds are used as sound sources in various species, but it is unknown how vocal fold morphologies are optimized for different acoustic objectives. Here we identify two main variables affecting range of vocal fold vibration frequency, namely vocal fold elongation and tissue fiber stress. A simple vibrating string model is used to predict fundamental frequency ranges across species of different vocal fold sizes. While average fundamental frequency is predominantly determined by vocal fold length (larynx size), range of fundamental frequency is facilitated by (1) laryngeal muscles that control elongation and by (2) nonlinearity in tissue fiber tension. One adaptation that would increase fundamental frequency range is greater freedom in joint rotation or gliding of two cartilages (thyroid and cricoid), so that vocal fold length change is maximized. Alternatively, tissue layers can develop to bear a disproportionate fiber tension (i.e., a ligament with high density collagen fibers), increasing the fundamental frequency range and thereby vocal versatility. The range of fundamental frequency across species is thus not simply one-dimensional, but can be conceptualized as the dependent variable in a multi-dimensional morphospace. In humans, this could allow for variations that could be clinically important for voice therapy and vocal fold repair. Alternative solutions could also have importance in vocal training for singing and other highly-skilled vocalizations.