<p>With the widespread application of computer processors, information processing technologies are rich and diverse in the digital information age. As an important research hotspot in signal processing, adaptive filtering algorithms have multiple application branches in signal processing. This study uses a deviation compensation term to infer and improve traditional adaptive filtering algorithms, thereby automatically updating filter parameters. Combined with different pulse noise and input signal noise environments, the adaptive filtering algorithms and input signal noise are analyzed and verified. By comparing the mean square deviation and performance of different algorithms, the cross-correlation adaptive filtering algorithm with a step factor of 0.4 and a kernel width of 4 for deviation compensation had the best convergence performance. At the same time, the mean square deviation of its algorithm was below − 22, indicating the superiority of the cross-correlation adaptive filtering algorithm for deviation compensation. Finally, comparing the mean squared deviation of different algorithms, the normalized minimum mean squared algorithm based on the Sigmoid architecture had the best mean squared deviation of − 33.9. The minimum mean squared algorithm based on affine projection-Wu-Manber least mean square had the best mean squared deviation of − 20.04. This indicates the superiority of the minimum mean square adaptive filtering algorithm based on deviation compensation. This is to suppress input signal noise interference and provide theoretical basis and technical reference for signal processing technology.</p>