Musical instruments recognition using hidden Markov model
Jonghyun Lee, Joohwan Chun · 2003
A new musical instrument recognition technique based on a hidden Markov model (HMM) is proposed. The spectral envelope is the key information of instrument characteristic and timbre. We decompose an instrument sound into sinusoidal components (harmonics) and noise components and estimate the amplitudes of the harmonics component. We want to express the spectral envelope effectively using estimated amplitude, therefore, we define three kinds of features and apply a recognition procedure to each feature. The HMM model used is continuous single Gaussian output HMM. To evaluate the performance of the recognition technique, the proposed technique is applied to classify the real instrumental sound of MUMS (MacGill University Master Samples). The recognition success ratio is more than 70%.