This project uses labeled audio files in order to build a fature-vector based model and be able to recognize some characteristics on a new unlabeled audio.

Pattern analysis is based on MFCC coefficients and GMM models.

Applications:

• automatic media indexing
• diverse contact center applications

Keywords:

• pattern recognition
• signal processing
• audio fingerprinting

Alternatively, neural networks, SVM, kD-Trees or other mechanisms can be applied to perform better recognition.