The Application of Speech/Music Automatic Discrimination Based on Gray Correlation Analysis

It is important for automatic discrimination of speech/music in content-based indexing and retrieval of cognitive multimedia. Previous work bases on longtime characteristics such as perceptual features like pitch, brightness, differential parameters, variances, time-averages of spectral parameters and etc. However, these algorithms are relatively complex and some of processing speed is not efficient. In this paper, the application of gray correlation analysis method in speech/music discrimination based on unique probability statistics of short energy root mean square (RMS) is present. The simulation results for different segments of music/speech signals discrimination and both algorithms indicate that gray correlation analysis is feasible. This new algorithm is more effective with accurate performance than that of the reference (C. Panagiotakis and G. Tziritas, 2005) in real-time multimedia applications