Improving Speaker-Independent Lipreading with Domain-Adversarial Training

08/04/2017

∙

by   Michael Wand, et al.

∙

0

∙

share

We present a Lipreading system, i.e. a speech recognition system using only visual features, which uses domain-adversarial training for speaker independence. Domain-adversarial training is integrated into the optimization of a lipreader based on a stack of feedforward and LSTM (Long Short-Term Memory) recurrent neural networks, yielding an end-to-end trainable system which only requires a very small number of frames of untranscribed target data to substantially improve the recognition accuracy on the target speaker. On pairs of different source and target speakers, we achieve a relative accuracy improvement of around 40 speech data. On multi-speaker training setups, the accuracy improvements are smaller but still substantial.