The Research And Implementation Of Speech Enhancement Algorithms Under Complex Conditions

With the development of science and technology,speech enhancement has important application prospect in vehicle system,video conference,hearing aid,smart home appliance,artificial intelligence and so on.In real life,due to the presence of noise,speech signal are disturbed and contaminated by different noises,which makes the speech quality very poor.Therefore,it is necessary to perform noise reduction processing on the noisy speech signal.Speec h enhancement technology can suppress the effects of noise,improve speech quality,and improve speech intelligibility and clarity.In the complex environment with diversity noise sources and environment reverberation,the combination of microphone array a nd speech enhancement technology can better reduce the impact of noise.In this paper,the speech enhancement algorithms have been analyzed and studied based on the complex environment.The main tasks are as follows:(1)The generalized sidelobe canceller(GSC)speech enhancement algorithm is analyzed and studied.The GSC speech enhancement algorithm is a classical beam former algorithm,and the GSC structure is simple and easy to implement.(2)The transfer function generalized sidelobe canceller(TF-GSC)speech enhancement algorithm has been analyzed and studied.The TF-GSC speech enhancement algorithm based on GSC speech enhancement algorithm,using the transfer function coefficient ratio to reconstruct the blocking matrix and fixed beamformer,compared with the GSC speech enhancement algorithm,TF-GSC speech enhancement algorithm improves the system's adaptability to the acoustic environment.(3)The convolutive function generalized sidelobe canceller(CTF-GSC)speech enhancement algorithm has been studied and improved.The CTF-GSC speech enhancement algorithm is based on the traditional GSC speech enhancement algorithm.It replaces the multiplicative transfer function(MTF)approximate form of the TF-GSC speech enhancement algorithm by using the CTF approx imate form,taking into account the mutual interference between different frequencies.A new structure of the blocking matrix and the fixed beamformer is constructed using the scale factor of the relative transfer function between the microphones,and a ne w array signal transfer function model is established using the CTF form of the frequency.The approximate form of the CTF is more accurate and less restrictive than the approximate form of the MTF estimate,further improving the noise reduction capability.The adaptive canceller module of CTF-GSC speech enhancement algorithm adopts the the normalized least mean square(NLMS)algorithm.The improved CTF-GSC speech enhancement algorithm uses a variable step-size-NLMS(VSS-NLMS)algorithm to construct an adaptive canceller module.The improved VSS-NLMS algorithm can further suppress the influence of input speech on noise,and make the adaptive canceller module have faster convergence and stronger stability,so that the CTF-GSC speech enhancement algorithm can better suppress noise.(4)The wiener filter speech enhancement algorithm has been studied in depth.The main principle of the traditional wiener filter speech enhancement algorithm is to minimize the mean square error between the desired speech signal and the estimated speech signal,so as to obtain a relatively pure speech signal.The improved wiener filter speech enhancement algorithm based on it,the voice activity detection(VAD)algorithm and the wavelet multi-window spectrum estimation algorithm are introduced to estimate the noise signal.The improved algorithm has less residual noise in the speech signal and better system noise reduction.(5)The speech enhancement algorithm based on CTF-GSC and post-Wiener filtering was analyzed and studied.In gener al,CTF-GSC speech enhancement algorithm has a good noise reduction effect on coherent noise,and wiener filter speech enhancement algorithm has a significant inhibitory effect on non-coherent noise.This paper combines these two methods effectively and pr oposes a new beamformer speech enhancement algorithm.The algorithm has obvious noise reduction ability for both types of noise and has good robustness.(6)In the complex environment,the paper carries the simulation analysis to the above algorithms,the speech signal acquisition and transmission experiment test.At the same time,the actual experimental environment is established and the algorithms are actually simulated.The experimental results are compared and analyzed.The experimental results show that the optimized algorithms proposed in this paper effectively improves the system signal-to-noise ratio and reduces speech distortion.In the complex environment with multiple noise sources and environmental reverberation,the proposed algorithm has better robustness.