Impact of feature selection and kernel functions in classification for MP3 bit rate detection.
Sen, Diana Qiong Ju.
Date of Issue2010
School of Computer Engineering
With the increased availability of audio files from the Internet, the quality of these audio files is of great concern for the audiophiles who download them. The bit rate of MP3 files is used to determine the audio quality. However, if the audio has been transcoded from lower bit rate to higher bit rate, it is unlikely to establish the true audio quality. Thus, investigation on how different kernel functions used in SVM to analyze different spectral density signal features will affect the classification results in determining the various bit rates with reference to an existing study deduction. This project will use three spectral density signal estimation methods namely, Pburg, Spectrogram and Periodogram to obtain feature data from three different audio datasets. Polynomial, radial basis function and sigmoid kernel functions were used in Support Vector Machine (SVM) learning based classifier to analyze the extracted features and recognize patterns with C-Support Vector Classifiers (C-SVC) for training and testing of data sets. In the absence of any coding format knowledge other than the audio frequency signal itself, the analysis returned an average success rate of 98.5% in correctly detecting the original compressed bit rate of an audio file. These success rates are very high with 99.49% also detected from the transcoding of lower bit rate, 128 kbps and 192 kbps to higher bit rate, 320 kbps.
DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing
Final Year Project (FYP)
Nanyang Technological University