The Research On Content-based Packet Loss Recovery For Classical Music

Packet loss is very common in voice signal processing. Many researchers at home and abroadare committed to study voice packet loss recovery, and also raised a lot of packet loss recoverymethods, but these repair methods are not concerned about the content of the speech signal itself.Music signal as a special class of the speech signal, its structure and the general speech signal is notthe same and they have their own unique property. Therefore some mature packet loss recoverytechnology is not fully applicable to the music signal. In such a background, the paper proposes acontent-based packet loss recovery method for classical music.In this article, we mainly discuss the packet loss recovery method for classical music.therefore we focus on the characteristics of the music content. The main work of this paper isdivided into the following steps: first, we extract containing the onset point of the data frame on thebasic of processing the music signal. Then, we extract the feature information of the data framecontained from the onset points. In this paper we extract the Mel-scaled Spectrum Coefficients,short for MFCC. Finally,we sort of these data frames containing the onset point according toMel-scaled Spectrum Coefficients and form a complete of packets including characteristicinformation to the receiving end.The key work of this article is to propose a specific signal loss recovery scheme according to thecharacteristics of the music signal and repair lost packet at the receiving end. We introduceclustering algorithm to classify feature information. Finally we evaluate our scheme using meanopinion score method. Experimental results show that the packet loss recovery method proposed inthis article is better than common recovery method.