The invention relates to a voice signal processing method, in particular to a voice enhancement method, which comprises the following steps of S10, audio acquisition, S20, screening effective voice signals, S30, preliminarily enhancing the voice, S40, screening target voice, S50, enhancing the voice signal again, S60, carrying out wake-up word detection: sending the re-enhanced voice into a high-precision wake-up word detection model, carrying out wake-up word detection, and entering S70 when a wake-up word is detected, and otherwise, returning to S20, and S70, continuously enhancing the voicedetected by the wake-up word, then sending the enhanced voice to a recognition end and carrying out recognition. According to the speech enhancement method provided by the invention, the calculationoverhead can be effectively reduced, and the recognition task can be accurately carried out even in a strong noise scene. The method is suitable for being applied to a local end with complex environment and limited computing resources.