Research On Audio Packet Loss Compensation

Network audio business is thriving with the worldwide popularity of the Internet and the development of digital signal processing technology, more and more high-fidelity audio is transported over Internet by compression coding. However, data packet switching technology based on IP protocol provides a "Best-effort service", and its transmission environment varies momentarily, which inevitably leads to packet delay and loss. It can't meet the requirement of real-time audio data transmission, resulting in a sharp deterioration in audio quality. Therefore, in order to improve the audio quality, it is extremely important that taking effective packet loss compensation technology predicts and estimates the information of the lost packet.Packet loss compensation can be viewed as the problem of estimating a lost frame, while using all available information, such that the generated output fits, as smoothly as possible, between the neighboring frames. The main work of this paper is:(1) Introduce prior work on Packet loss compensation technology.(2) Briefly present the kernel of perceptual audio encoder, mainly focus on the principle of MPEG-2 AAC, including the psychoacoustic model, filter bank, quantization coding, joint stereo and temporal noise shaping module.(3) Describe the basic theory and knowledge we use in our packet loss compensation algorithm, including the sinusoidal plus noise model, MDCT transform theory, the MDCT analysis of sinusoidal signal and pseudo-spectrum.(4) Structure the framework of our packet loss compensation and introduce each module.(5) Carry out algorithm simulation in MPEG-2 AAC standard, the algorithm proposed in this paper is compared with existing algorithms through PEAQ and MUSHRA test.