Connectionist speaker normalization and its applications to speech recognition

Speaker normalization may have a significant impact on both speaker-adaptive and speaker-independent speech recognition. In this paper, a codeword-dependent neural network (CDNN) is presented for speaker normalization. The network is used as a nonlinear mapping function to transform speech data between two speakers. The mapping function is characterized by two important properties. First, the assembly of mapping functions enhances overall mapping quality. Second, multiple input vectors are used simultaneously in the transformation. This not only makes full use of dynamic information but also alleviates possible errors in the supervision data. Large-vocabulary continuous speech recognition is chosen to study the effect of speaker normalization. Using speaker-dependent semi-continuous hidden Markov models, performance evaluation over 360 testing sentences from new speakers showed that speaker normalization significantly reduced the error rate from 41.9% to 5.0% when only 40 speaker-dependent sentences were used to estimate CDNN parameters.<<ETX>>