The Study On Speech Enhancement Technology Of Omni-directional Microphone Based On Improved Wiener Filtering Algorithm

Language is the bridge of communication between human beings.Speech is the voice of the language.It is the most direct and natural way for human to exchange information and communicate with others.With the progress of the Times,mankind has stepped into the era of information,and Using modern digital technology to solve the storage,processing and transmission of speech signals is the main task of speech signal processing.However,in the real life environment,noise exists all the time,which seriously affects the intelligibility of speech signals.Therefore,it is an important research subject for speech signal processing to eliminate the noise,achieve the enhancement of speech and improve the stability of speech processing system.In this paper,two main speech enhancement algorithms based on short-time spectrum estimation are studied,including spectrum subtraction and Wiener filter algorithm.In addition,the two algorithms are improved,and the speech enhancement effect is evaluated by subjective evaluation and objective evaluation respectively.The results show that the denoising effect of the Wiener filtering algorithm based on the prior Signal-to-noise rate is much better than that of Spectral subtraction,but there is still residual noise.Therefore,it is decided to improve the noise estimation method based on the wiener filtering algorithm based on the prior SNR.Traditional noise estimation is based on the calculation of the average value of the power spectrum of preamble silent segment,which will lead to noise residue and bad effect of noise reduction.In this paper,on the basis of the wiener filtering algorithm based on the prior SNR,by using the correlation between speech frame to estimate the probability of pure speech signals in each frequency point.Then,the noise power spectrum will be calculated,which can be updated in real time.The improved Wiener filtering algorithm has a good effect on nonstationary noise.For the actual conference site,the volume is too small due to the fact that the voice of the living person is small or the distance between the speaker and the microphone is a bit far away,then the automatic gain volume control algorithm is programmed,which can effectively increase the pickup distance of the conference call to 3 meters.The hardware architecture of the omnidirectional microphone designed in this paper is based on TI company's DSP(TMS320C6748),ST company's ARM(STM32F429)and TLV320AIC3104 audio coding-decoder(Codec)as the core of embedded hardware platform.With the DSP's efficient digital signal processing performance,the speech denoising algorithm can be achieved.ARM is mainly for the convenience of human-computer interaction,and facilitates the expansion of the later development.For example,by the communication of audio data between ARM and Andriod,and the development of voice APP can achieve the communication of the IOT conference call.