Sound Effects Classification and Retrieval
The core of the fine-level classification and retrieval of environmental sound (including sound effects) is to build the hidden Markov model (HMM) for each class or clip of sound(s). Currently, two types of information are contained in the model, i.e. timbre and rhythm. Each kind of timbre is modeled as one state of the HMM, and represented with the Gaussian mixture density. The rhythm information is included in the transition parameters of HMM. For sound effects classification, HMM serves as the classifier; while for sound effects retrieval, it is the similarity measure.
KeywordsFeature Vector Hide Markov Model Gaussian Mixture Model Viterbi Algorithm Sound Effect
Unable to display preview. Download preview PDF.