Estimation of channel bias for telephone speech recognition

We propose a maximum a posterior (MAP) estimation of channel bias to compensate for the channel mismatch in telephone speech recognition. For a telephone speech, the channel bias is estimated by maximizing a posterior probability. Because a posterior probability is composed of a likelihood function and a prior density, we introduce a scale factor to evaluate their weights in MAP estimation. To further improve the performance, a priori channel statistics is extended to multiple components and the channel mismatch is separately compensated for different segments. A rapid MAP estimation applied in the feature domain is also proposed for reducing computational complexity. Experiments show that the proposed method can significantly improve recognition rates and computational complexity