Detection Of Weak Speech Signal In Strong Noise Background Based On Random Resonance Theory

In our daily life, people transmit information or communicate with each other and so on mainly through voice, so the voice is the most commonly used, convenient and effective way of communication by people. During communication, however, the speech signal inevitably will be affected by the internal or external noise interference, and then what we received is the speech signal with noise, rather than the pure original speech signal, which will disturb people in receiving information quickly and accurately. So, the scholars try to extract the purest speech signal from the noise speech signal by speech enhancement, and improve its quality in order to analysis the speech signal effectively.According to the difference of the noise characteristics, the scholars have proposed many speech enhancement methods. At present, some of the more commonly used methods are spectrum subtraction, adaptive noise-suppression, wavelet transformation and Wiener filter and so on. But the methods mentioned above mainly treat the noise as harmful interference signal, and eliminate the noise from the noise speech signal by noise estimation. While under the background of strong noise, the de-noising effects get worse, as they not only remove the noise but also lost part of the speech information, or make the waveform distortion.Stochastic resonance is a method of which can transfer the noise energy to weak signal, and thus it amplifies the weak signal and suppresses the noise. According to stochastic resonance theory, a new method based on adaptive stochastic resonance to extract weak speech signals is proposed. This method, combined with scale-transformation, realizes the detection of weak speech signals from strong noise. By evaluating the signal-to-noise ratio of the output signal, we can adaptively adjust the parameters of the system, and then the weak speech signal is optimally detected. Experimental simulation analysis shows that under the background of strong noise, the output signal-to-noise ratio increases compared to the initial value. This method obviously raises the signal-to-noise ratio of the output speech signals, which gives a new idea to detect the weak speech signals in strong noise environment.