Abstract: It increases the risk of irreversible hearing loss for long-term loud listening. Therefore, in order to better prevent the hearing loss caused by long-term loud listening, this paper implements an active ear-protecting ear-protecting voice processing accelerator based on the optimized multi-channel wide dynamic range compression (WDRC) algorithm. Compared with the traditional WDRC algorithm, the accelerator takes the input audio sound pressure level as the input parameter of the WDRC algorithm, and compresses the input audio. The input voice is divided into 16 channels according to the Mel scale. The input voice that is too loud in each channel will be compressed. The compressed voice signal is combined by a comprehensive filter, processed by the ΔΣ modulator and output in the form of pulse density modulation (PDM). The test results show that for voice input with higher sound pressure levels in different frequency bands, a better sound pressure level control effect is produced, and real-time control of the output voice sound pressure level is realized.