Estimating multiple pitch frequencies of concurrent speech sources from a single-microphone input is essential to speech separation. Nevertheless, pitch cues of individual sources are weakened by each other, making the estimation unreliable. This paper presents a pitch tracking method that incorporated in a model-based separation framework. Multiple pitch estimation is simplified into single pitch estimation by segregating the source envelope from mixture spectrum with statistics of familiar speech patterns. Comprehensive experiments have compared the proposed tracking method with a recently reported multiple pitch estimator and its modified version equipped with ideal pitch cues. Lower estimation errors are achieved. Furthermore, this approach is applicable to other model-based frameworks as well.

No abstract

No abstract

This paper presents a set of eective and ecient techniques to improve the discrimination capability of a recurrent neural network (RNN) based isolated word recognizer. The recognizer contains a set of individually trained RNN speech models (RSMs). Each of them represents a dierent word in the vocabulary. Speech recognition is performed by selecting the RSM that best matches the input utterance. For temporal supervised training of the RSMs, a new error function is introduced, in which the contributions of all phonetic components are equalized regardless of their dierence in duration. The learning rate for recurrent connections is amplied. This is aimed at strengthening temporal dependency in the RSMs to capture dynamic characteristics of speech signals. Furthermore, a hierarchical training strategy is employed to facilitate more efcient discriminative training among the RSMs. A series of speaker-dependent recognition experiments are performed to evaluate the eectiveness of the proposed techniques.

No abstract