Arithmetic Research On Audio Signal Restoration

In the public security system, the identification of audio plays an important part in the work of legal organs. The recordings of dialogues during calls, interviews and the covert surveillances, as well as the texts of transcribed recording, are the key evidences. However, owing to the existence of kinds of noises outside, the intelligibility of those audio evidences are often very low, thus decrease the credibility. Based on the advanced signal processing arithmetic, audio restoration aims at removing the useless parts from the mixed audio signal to enhance the intelligibility of the audio signal.This paper introduces the basic knowledge of audio signal, as well as the basic principles of linear prediction, makes a detailed description about a least square AR (LSAR) repairing algorithm based on AR model and the theory of minimum mean square error, and investigates on impulse noise cancellation and clipping restoration of the multi-module noise processing platform in audio signal. In the module of impulse noise cancellation, a detection algorithm based on the residual domain of the audio signal is proposed for the characteristics of impulsive noise, and then the simple straight-line interpolation and LSAR arithmetic are utilized to restore the audio. In the module of clipping restoration, a detection arithmetic based on the peak value of audio's amplitude is proposed through studying the characteristics of clipping distortions, and then the LSAR algorithm is used to finish the audio restoration. With the analysis of simulation results and the subjective and objective evaluation based on the value of the perceptual evaluation of speech quality (PESQ) and the signal to noise ratio (SNR) on the corresponding test suite, the repairing arithmetic is effective to the degraded audio, and improves the intelligibility of the audio.