Multi-resolution Spectral Entropy Based Feature for Robust ASR

Recently, entropy measures at different stages of recognition have been used in automatic speech recognition (ASR) task. In a recent paper, we proposed that formant positions of a spectrum can be captured by multi-resolution spectral entropy feature. In this paper, we suggest modifications to the spectral entropy feature extraction approach and compute entropy contribution from each sub-band to the total entropy of the normalized spectrum. Further, we explore the ideas of overlapping sub-bands and the time derivatives of the spectral entropy feature. The modified feature is robust to additive wide-band noise and performs well at low SNRs. In the last, in the frame work of TANDEM, we show that the system using combined entropy and PLP features works better than the baseline PLP feature for additive wide-band noise at different SNRs.