Abstracts Submitted: 510
Number of Users: 657
View Abstracts Submitted
Back to home Page
In this paper, we propose a method of harmonic enhancement for speech acquired by a laser microphone. It is difficult to acquire the distant sound source because an air-conduction microphone also acquires unnecessary sounds near the microphone. This is because it acquires the sound which reaches the diaphragm regardless of the distance. A laser microphone is studied as the solution of this problem. It can measure the vibration of a distant object by ir-radiating the laser beam to the surface of the distant object. For this reason, it can also ac-quire the distant sound source which reaches the distant object. In this study, we use a laser Doppler vibrometer (LDV) as the laser microphone. However, in the sound acquired by the laser microphone, the high frequency is attenuated and the sound is collapsed by the high-frequency noise. Because of this, speech becomes unclear by losing harmonic structure which is included in the speech characteristics. Previous studies of noise reduction have spectral sub-traction (SS) and iterative spectral subtraction (I-SS). SS reduces the noise by subtracting the estimated noise in frequency domain. I-SS is the improved noise reduction to suppress the musical noise which is the residual noise left after SS. However, it has the problem that the processed speech was distorted, since it reduces speech along with the noise. We focus on that harmonic structure of speech appears in the harmonic tones of the fundamental frequency. The proposed method implements the enhancement of harmonic structure by modifying I-SS in the harmonic tones of the fundamental frequency. To solve high frequency attenuation, it has the pre-emphasis filter that increases the power of high frequency to fix the bias in frequency domain. We carried out an evaluation experiment to verify the effectiveness of the proposed method. In this evaluation, we acquired speech by the laser microphone by changing the object irradiating the laser beam. We compared the sound quality of clean speech, and speeches to which we applied the previous method and the proposed method. As a result of the evaluation experiment, it was found that the proposed method is more effective than the previous method.
© Copyright 2017 All Rights Reserved