Channel and Noise Robustness of Articulatory Features in a Deep Neural Net Based Speech Recognition System

Abstract

Articulatory features (AFs) are known to provide an invariant representation of speech, which is expected to be robust against channel and noise degradations. This work presents a deep neural network (DNN)—hidden Markov model (HMM) based acoustic model where articulatory features are used in addition to mel-frequency cepstral coefficients (MFCC) for the Aurora-4 speech recognition task. AFs were generated using a DNN trained layer-by-layer using synthetic speech data. Comparison between baseline mel-filterbank energy (MFB) features, MFCCs and fusion of articulatory feature with MFCCs show that articulatory features helped to increase the noise and channel robustness of the DNN-HMM acoustic model, indicating that articulatory representation does provide an invariant representation of speech.

Topics: Speech communication, Acoustics, Speech recognition, Auroral phenomena, Artificial intelligence, Artificial neural networks, Signal processing, Markov processes