Research On Audio Classification Technology

Speech and music are the most important of the two types of audio information. Audio information processing plays an important role in multimedia information processing. Because the characteristics of the audio information itself, audio classification techniques for audio-structured has a critical role of the audio information and is directly related to the in-depth analysis of the extent and degree of accuracy, as well as the extraction of semantic audio content.This major work and research findings include the following aspects:(1) The idea and realization block diagram of combined speech and audio coding were introduced, as well as a variety of classifier and classification algorithms design. An method of audio features analysis and extraction with audio-clip and audio-frame was analysed in detail. And two kinds of typical speech and audio codec: Code Excited Linear Prediction (CELP) and Advanced Audio Coding (AAC) were briefly described.(2) Two kinds of existing audio classification algorithm: audio classification algorithm based on Linear Prediction Coding (LPC) and pitch intensity were analysed in detail. And their computational complexity and classification accuracy were compared by the experiments.(3) A real-time audio classification algorithm based on the tonal characteristics was presented by analysing and extracting tonal characteristics (the number of tone, low tone frame rate or the number of sub-band tone ratio) and time-domain characteristic parameters ( zero-rate or spectral tilt) of the audio signal using psychoacoustic model 1. Designed a variety of implementation methods and contrasted their the classification performance to obtain the best method: classifying speech and music using the number of sub-band tones ratio and spectral tilt with smoothing for the classification results.(4) The classification performance of the audio classification algorithm proposed in this paper and the existing algorithms were compared by experiments.The experiment results showed that the proposed audio classification algorithm effectively reduced the computational complexity and improved the classification accuracy and achieved real-time classification compared to the existing methods. Therefore, the proposed algorithm was more practical than the existing methods.