English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

When using ultrasound video as input, Deep Neural Network-based Silent SpeechInterfaces usually rely on the whole image to estimate the spectral parametersrequired for the speech synthesis step. Although this approach is quitestraightforward, and it permits the synthesis of understandable speech, it hasseveral disadvantages as well. Besides the inability to capture the relationsbetween close regions (i.e. pixels) of the image, this pixel-by-pixelrepresentation of the image is also quite uneconomical. It is easy to see thata significant part of the image is irrelevant for the spectral parameterestimation task as the information stored by the neighbouring pixels isredundant, and the neural network is quite large due to the large number ofinput features. To resolve these issues, in this study we train an autoencoderneural network on the ultrasound image; the estimation of the spectral speechparameters is done by a second DNN, using the activations of the bottlenecklayer of the autoencoder network as features. In our experiments, the proposedmethod proved to be more efficient than the standard approach: the measurednormalized mean squared error scores were lower, while the correlation valueswere higher in each case. Based on the result of a listening test, thesynthesized utterances also sounded more natural to native speakers. A furtheradvantage of our proposed approach is that, due to the (relatively) small sizeof the bottleneck layer, we can utilize several consecutive ultrasound imagesduring estimation without a significant increase in the network size, whilesignificantly increasing the accuracy of parameter estimation.

Autoencoder-Based Articulatory-to-Acoustic Mapping for Ultrasound Silent Speech Interfaces