Real-time speech processing on pitch estimation and speech enhancement using binaural information
禎史 苣木 · Kumamoto University Repository (Kumamoto University) · 2007
This dissertation addresses important issues in fundamental speech processing technologies for a hearing assistance system with spatial selectivity. Two major topics are mentioned. One is pitch detection in a noisy environment and the other one is speech enhancement with spatial selectivity. This dissertation proposes new algorithms for both topics and discusses real-timeprocessing. Chapter 2 introduces a pitch detection algorithm based on the harmonic wavelet transform. A speech encryption algorithm is also proposed as one of pitch-based applications. In chapter 3, a parallel algorithm for pitch detection in real-time processing is mentioned. Furthermore, a pitch detection system under concurrent speech condition is proposed. Chapter 4 describes a sound source localization and segregation algorithm in azimuth and elevation using binaural information. Chapter 5 addresses the real-time processing of a binaural model. Finally, chapter 6 concludes this thesis.