Acoustic signal-based approach for fault detection in motorcycles using chaincode of the pseudospectrum and dynamic time warping classifier

The sound of a moving vehicle gives a clue of the fault. This study investigates fault detection of motorcycles using chaincode of the pseudospectrum. The motorcycle sound signals are analysed for spectral variations and these variations are traced by a chaincode. The chaincode features are used to classify the sample into healthy or faulty using dynamic time warping technique. MATLAB version 7.8.0.347 (R2009a) is used for effective implementation. The classification results obtained are over 91% and 93%, respectively, for faulty and healthy motorcycles. The results are comparable with the reported works based on wavelets. The proposed work finds applications in traffic census, traffic rule enforcement, machine fault discovery, automatic surveillance and the like.