Research On Transient Noise Suppression Algorithm In Speech System

The speech is one of the main ways of human information exchange,but there are many different noises in our living environment.When we use the communication equipment for speech communication process,people are increasingly concerned about the quality of speech in order to ensure the accuracy of information transmission.Therefore,the most important performance of the speech system is the property to suppress noises.At present,there are many ways to achieve a successful suppression of the stationary noises such as gaussian white noise.However,there are many common transient noise in our life.It can also seriously affect the speech system performance.Such as knocking,applause,electromagnetic interference,etc.Transient noise has a common feature is the sudden strong,it was oscillating in the form of attenuation in the time domain,the duration of more than ten milliseconds to hundreds of milliseconds,it is very widely distributed in the frequency domain.So the frequency spectrum of the transient noise is basically mixed with the normal voice spectrum in the frequency domain,it is difficult to suppress the transient noise.Therefore,this paper mainly researches and realizes the suppression algorithm of transient noise in speech system.In this paper,I use two algorithms to research the problem of transient noise suppression.One is Optimally Modified-Log Spectral Amplitude(OM-LSA)algorithm,which is for the general class of transient noise.It needs to estimate the priori signal to noise ratio and the probability of speech,and the transient noise spectrum is estimated by using the improved minimum value control recursive average(MCRA)algorithm.Finally,the spectral optimal gain function and the optimal logarithmic spectrum is obtained.I use IFFT to derive the pure speech signal.The other is Energy Normalized Variance(ENV)algorithm,which is for the impulse noise.The impulse noise is detected in the time domain and the impulse noise is suppressed by the amplitude threshold.In this paper,I use the transient noise suppression algorithm for different types of speech segment,which are disturbed by a variety of transient noise.I Analyze and test all kinds of transient noise in time domain and frequency domain with Cool Edit Pro audio processing software and phonological diagram.The performance of the algorithm was evaluated by the segmentation signal-to-noise ratio(Seg SNR)and the average subjective score(MOS).The results show that the algorithm proposed in this paper can improve the speech quality effectively by suppressing the transient noise.