Sound Quality Discrimination of the Impulsive Sounds with Mahalanobis Distance

Abstract

Reduction of the acoustic noise emitted by a copier and printer is strongly demanded because of more rigorous noise regulations and higher customer requests. Although to identify the noise sources is important, the conventional technologies to search for the locations of noise sources with phase information do not have enough spatial resolution for the impulse noise. So an analysis method to discriminate between impulse sounds with sound quality characteristics have been studied for identification of noise sources. As the results, the following conclusions were gotten. (1) A new method to discriminate between impulse sounds by utilizing a statistical analysis of sound quality was devised, which directly expands each data on the acoustic waveform to Mahalanobis space and compares spatial distances. (2) It was shown that four types of simulated impulse sounds emitted from paper transport systems could be discriminated with accuracy of 90%. (3) It was clarified that how to normalize waveforms on the time axis was important to improve the discrimination accuracy. (4) Mahalanobis distance has continuous and monotonous characteristics corresponding to the differences between waveforms.