Noise and various disturbances in the high voltage and strong magneticfield environment directly impact the accurate detection of electrical signals, as well as interfere with the time-frequency characteristics of these signals. To mitigate noise and disturbance components, an adaptive noise reduction method is proposed based on wavelet frequency division threshold and frequency division threshold function. Firstly, following wavelet packet decomposition of the signal, the node coefficients from the final two layers of the wavelet packet tree are arranged in order of frequency magnitude, and the node coefficient thresholds for different frequency bands (frequency division thresholds) are estimated. Secondly, the noise reduction adjustment coefficient is constructed by the ratio of energy of adjacent layer node coefficients, and the frequency division threshold is adaptively optimized. Subsequently, an improved threshold function (frequency division threshold function) is constructed by incorporating the benefits of both soft and hard thresholds, utilizing the noise reduction adjustment coefficient to achieve adaptive adjustment of threshold functions of different nodes. Finally, the denoised signal is obtained by reconstructing the wavelet packet tree. Simulink simulation and noise reduction results of the measured recorded waveform signal under high-voltage strong magnetism show that the method can effectively remove some effects, such as spikes, burrs, etc. The noise reduction effect is significant and the signal waveform restoration is excellent. The method holds certain application value in electrical signal detection.