Research On Improved Algorithm About Speech Enhancement

In practical applications, such as speech telecommunication, speech synthesis,speaker recognition, the degradation of speech by all kinds of noise is inevitable and itaffects the performance of the speech equipments very badly. So, it is necessary todenoise for the degraded speech, this is called SPEECH ENHANCEMENT.The approaches of speech enhancement are numerous. The first part of this paperis the origin and history of speech enhancement, several short-time spectrumestimation enhancement approaches and some knowledge related to speechenhancement are given too, for example, the models of speech and noise, and theestimation of noise.Generally speaking, a noise estimation algorithm is a noise variance estimator.The accuracy of noise variance estimation will affect next process. A noise estimationbased on minimum statistics is deeply studied in this thesis. The evaluation of themethod is given after a simulation experiment. The speech enhancement system ofthis thesis will adopt the noise estimator based on minimum statistics.The goal in post proc.essing in the thesis is to improve quality of speech, reducedistortion of the speech and decrease the effect of "musical noise".A new method based on masking properties of human ear and improvedspectrum estimation for speech enhancement is proposed. Simulation result showsthat the performance of the modified approach is better than the original one. Themasking effect is used in post processing, experiment results show that speech qualityis improved.