The Speech Enhancement through Elimination of Impulsive Disturbance Using Log MMSE Filtering

Abstract

The purpose of speech is communication, i.e., the transmission of messages. A message represented as a sequence of discrete symbols can be quantified by its information content in bits, and the rate of transmission of information is measured in bits/second (bps). In speech production, as well as in many humanengineered electronic communication systems, the information to be transmitted is encoded in the form of a continuously varying (analog) waveform that can be transmitted, recorded, manipulated, and ultimately decoded by a human listener. In the case of speech, the fundamental analog form of the message is an acoustic waveform, which we call the speech signal. Speech signals can be converted to an electrical waveform by a microphone, further manipulated by both analog and digital signal processing, and then converted back to acoustic form by a loudspeaker, a telephone handset or headphone, as desired. Signals are usually corrupted by noise in the real world. To reduce the influence of noise, two research topics are the speech enhancement and speech recognition in noisy environments have arose. It provided that better results in terms of performance parameters, processing time and speech signal quality rather than prior methods.