We have proposed a speech enhancement method based on formant level control, where a few kinds of formant components are enhanced individually and summed to yield an enhanced speech signal. As one of practical applications, we have been developing a speech signal processing to recover formant levels of the broadcasting speech to be radiated in a noisy place. In such a processing, the speech formant components are controlled according to frequency responses of the background noise before radiating, and then the enhanced speech is radiated in the noisy place. As a preliminary test, we have carried out an articulation test using VCV phonemes and succeeded in maintaining the high articulation scores under the bubble noisy condition of about SNR=-15dB. In addition, the proposed method will be applied in the area of speech processing for hearing aids. Especially, it is expected to apply not only in a compression type of hearing aids but also in a binaural hearing aid, which transmits the selected formant components separately to each of user's ears, according to a principle of formant decomposing and enhancing. In this paper, we describe a principle of the formant based speech enhancement and the experimental results of speech perception with noise, and then show some of the confirmed results by the preliminary investigations that this speech enhancement method will be effective for such a new hearing aid.
展开▼
