Multi-channel frequency domain speech enhancement algorithm based on variable-tension generalized subspace

Abstract

The invention discloses a multi-channel frequency domain speech enhancement algorithm based on a variational generalized subspace, and the algorithm comprises the steps: converting the data of a time domain into a frequency domain, extracting a generalized feature vector of an updated signal covariance matrix through a generalized subspace tracking algorithm, and constructing a variational filter; and the filter carries out filtering processing on different sub-bands of frequency domain data, the filtered signal obtains a statistic similar to that of a pure voice signal, and a good filtering effect is obtained. The algorithm has certain expansibility, can coordinate the balance between the voice output signal-to-noise ratio and voice distortion, and can be applied to real-time voice noise reduction processing occasions.

Description

technical field [0001] The invention belongs to the fields of adaptive signal processing and speech signal processing, and in particular relates to a multi-channel frequency-domain speech enhancement algorithm based on variable expansion into a generalized subspace. Background technique [0002] Speech signals are an important means of communication between living things. With its rich information content, speech signal has become the most effective means of communication between individuals. The development of voice signal technology began with Bell's invention of the telephone in the second half of the 19th century. The invention of the telephone has brought human communication ability to a higher level, greatly improved the efficiency of social operation, and set off the trend of information science and technology development. The processing of speech signals usually uses acoustic sensors, and the virtual acoustic energy is transformed into quantitative electrical signa...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome above-mentioned deficiency, provide a kind of multi-channel frequency-domain speech enhancement algorithm based on variable tension into generalized subspace, solve the problem of coordination between speech signal output signal-to-noise ratio and speech distortion, and provide a kind of Online Efficient and Fast Speech Enhancement Algorithm

Images