Research On Key Technology Of Noise Reduction Headphones In Factory Environment

The main way for people to communicate with each other is through speech,and the biggest interference to speech signals is noise,which makes speech unclear and reduces its comprehensibility.The absence of location information also interferes with judgement,affecting normal communication.Especially in a factory environment,noise interference and the absence of location information create inconvenience and hidden dangers for workers.Therefore,in a factory environment,noise reduction needs to rely on better speech noise reduction technology while also ensuring that location information in speech is not lost,in order to maximize the efficiency and safety of workers’ use of speech.This article aims to improve the noise reduction performance of noise-canceling headphones in a factory environment,improve speech quality,and perform sound source localization.To this end,this article starts from speech endpoint detection and proposes a speech endpoint detection method based on Empirical Mode Decomposition(EMD)and Teager energy algorithm to improve the autocorrelation function.To address the shortcomings of inaccurate endpoint detection resulting in poor enhancement performance in the Wiener filtering speech enhancement algorithm,the detection process has been improved based on precise endpoint detection,enhancing the enhancement effect.To solve the problem of lost location information in speech after noise reduction,this article uses speech synthesis reconstruction technology to achieve spatial audio source localization.The work and innovation of this article are as follows:(1)Research the current status and development trends of speech enhancement technology,introduce speech preprocessing technology,and focus on the characteristic analysis and classification of noise in factory environments.Additionally,provide a detailed introduction to methods of evaluating the quality of speech signals to evaluate the processed speech signals.(2)To address the problem of difficulty in accurately detecting speech endpoints in the harsh noise environment of factories,a speech endpoint detection method based on the improved autocorrelation function for EMD and Teager energy algorithms is proposed.Using multi-window spectral estimation and spectral subtraction for noise reduction,the empirical mode decomposition is applied to improve the signal-to-noise ratio.Then,the noisy intrinsic mode function data sequences are discarded and appropriate components are selected for speech reconstruction.The Teager energy operator of each order of IMF component is calculated,and endpoint detection is performed using a logarithmic energy and correlation function method.Experimental results show that the proposed endpoint detection method has significantly improved detection accuracy in factory environments.(3)To address the problem of inaccurate estimation of prior noise segment power spectra and amplitude spectra in the classic Wiener filtering speech enhancement algorithm,an improved Wiener filtering speech enhancement algorithm is proposed using the precise endpoint detection method.Experimental results show that the improved Wiener filtering speech enhancement algorithm has good noise reduction effects in factory environments.(4)To address the problem of lack of positional information when converting mono speech signals into stereo spatial audio after noise reduction processing,a speech synthesis and reconstruction model is designed for factory environments.This model extracts binaural loudness difference clues from the original speech signals,reconstructs the spatial audio speech signals,and analyzes their perceptibility at different horizontal azimuth angles.