The Application of Spectral Subtraction and Calman Filtering for Audio Noise Reduction

Spectral subtraction is a spectrophotometric determination method for content measurement that eliminates interfering components through mathematical processing. It obtains a pure signal by subtracting the spectrum of noise. If we assume that the noise is additive and, through short-time spectral analysis, considers the spectrum to be stable over short periods. By calculating the spectrum of noisy speech and the noisy spectrum, the spectrum of pure speech can be obtained. Spectral subtraction is widely used in the field of speech denoising due to its simplicity and fast processing speed. For example, in real-time voice communication, spectral subtraction can effectively remove background noise and improve speech quality. The main drawbacks of spectral subtraction include the generation of "musical noise", which may lead to rough speech quality. Therefore, further methods such as hybrid spectral-Kalman filters have been proposed by combining Kalman filtering techniques, achieving better performance. This paper reviews and compares the application of spectral subtraction and Kalman filtering for noise reduction in complex noise environments, and points out future research challenges and directions.