Abstract: The performance of traditional spectral subtraction algorithms is often unsatisfactory in the case of low signal-to-noise ratio conditions. To solve this problem, an improved spectral subtraction algorithm based on the Katz dimension is proposed. Furthermore, a new strategy involving the dynamic adjustment of over-subtraction factor α is used to improve spectral subtraction, which improves the algorithm's performance. Experimental results show that compared with the traditional spectral subtraction, wavelet transform, and spectral subtraction combined with wavelet transform speech enhancement method, the proposed algorithm improve the results in three aspects: signal-to-noise ratio, perceptual evaluation of speech quality, and short-time objective intelligibility.