The Research On Speech Dereverberation

When a microphone is placed too far from a speaker, for example, in hands-free telephone, hearing aid and tele-conferencing systems, people often suffer from reverberation indoor. Strong reverberation makes pinches of speech overlap each other, and furthermore results in degradation in speech. So, dereverberation becomes one of the main tasks in speech signal processing.Speech dereverberation (SDR) is an important part in transparent acoustic communication (TAC), which is also a key technology for speech enhancement. Simultaneously, it provides pre-processing for many speech signal processing such as speech source localization and speech recognition.SDR technology is also widely used in many acoustic fields such as multipath propagation, architecture acoustics, ocean acoustics, seismic data analysis, ultrasonic measure, biomedicine, vibration acoustics, radar and sonar.The research on dereverberation in China mainly focuses on ocean reverberation, where the sources used are often short time impulses or single tones. While in foreign countries, more research is on non-stationary speech signal dereverberation, which becomes a hot field in the last ten years.This thesis is arranged as follows:1. The principle of reverberation and its characteristics and mathematic model are first introduced. A lot of dereverberative applications, especially including speech signal processing, are summarized.