Design and implementation of synthesis algorithm for Chinese speech-like noise

Masking the speech signal by using noise is an effective method to protect speech privacy. Speech-like noise has a better masking effect than other natural or artificial noise signals. In this paper, we propose a random probability method to generate speech-like noise. The method generates random text according to a random number and the probability table of Chinese language, and then inputs the text to speech (TTS) software to synthesize speech-like noise based on the random text. The speech-like noise has good masking properties and anti-cracking ability. The test results show that when the signal-to-nose ratio is less than or equal to -4dB, the speech intelligibility is less than 45%.