Brief Review of Recent Researches in Speech Enhancement from Filters to Neural Networks

According to the World Health Organization, more and more people will suffer from hearing loss in the future. Therefore, there will be greater demand for the output and technology of hearing aids. Under the help of artificial intelligence, the technology of smart hearing aids will also become more intelligent, so that the wearer can get a better experience. This paper mainly studies the related problems of speech signal processing in intelligent digital hearing aids. This article focuses on the speech enhancement in digital hearing aids. First, this work studies the application of filter in speech enhancement technology, mainly introduces the Wiener filter algorithm using the minimum mean square error criterion; the Kalman filter algorithm that can solve discrete signals; spectral subtraction based on multi-window spectrum estimation. Secondly, this paper studies some specific methods of deep learning technology in speech enhancement, mainly introduces DNN-based speech enhancement method, deep learning-based auditory cepstrum coefficient speech enhancement algorithm, and AE-CGAN-based speech enhancement algorithm. Finally, this paper studies the related problems of acoustic scene classification, divides the listening environment into Gaussian white noise, impact noise, and music noise, and explains the solutions for each listening environment.