| Summary: | 碩士 === 國立成功大學 === 系統及船舶機電工程學系 === 102 === Nowadays, cellular phones which are used almost everywhere for communication around the world become more and more popularly and the whole performances of this device are highly improved no matter in communication, image capture, and human machine interface except for background noise reduction ability. Until now, in a noisy environment, the communication quality is still awful because the acoustic environments where people talk are quite complicated and there exists no suitable solution for it. Based on the above reasons, this research proposes three new noise reduction methodologies to solve the background noise corrupted problem of cellular phone based on optimal and robust estimation concepts. In the hardware configuration, two different type microphones are used: the first one is an omni-directional microphone embedded in rear of the cellular phone for all sound sources collection purpose, and the other (uni-directional microphone) is arranged in front of the cellular phone directly closely to the user’s mouth. Collecting signals of these two microphones are whiten first as a speech signal corrupted with a background noise which can be regarded as a white noise and mathematically modeled by recursive least square (RLS) system identification then. Based on these prior arrangements, three estimators: Kalman filter, H2 estimator and H∞ estimator, are further developed to eliminate the unwanted environment background noises. From the performance tests, obvious background noise reduction capabilities for the speech enhancement of communication quality of cellular phones are delivered in this investigation.
|
|---|
