Sudha Rani Narasimhan
1540 Neston Way, Los Altos, CA 94024
The objective is to identify musical instruments using a computer. The characteristics of a musical instrument can be determined by the frequency spectrum of a specific note. The spectrum will vary for different instruments. This variation identifies the instrument. The waveforms of musical instruments (simulated by an electronic keyboard) are stored in the computer. Experimental data were collected for clarinet, trumpet, flute, accordion, violin, and piano. A computer program is developed to characterize the various spectra. The frequency of maximum power is determined first. Then the ratios of the power levels corresponding to the first five harmonics are calculated. For each instrument, the above ratios are determined from several trials. It is found that a characteristic feature of a musical instrument corresponds to specific values of these ratios. A computer program compares the ratios obtained from playing an instrument in real time with the stored ratios of various instruments to identify and display the instrument played. The identification method agrees very well with psychoacoustic data obtained from human beings [Saldanha and Corso, J. Acoust. Soc. Am. 36, 2021--2026 (1964)].