J. Acoust. Soc. Am., Vol. 120, No. 5, Pt. 2, p. 3323, 2006 (Invited Paper)
Preprocessing speech against reverberation
T. Arai
Abstract: Although short reverberation in a room may help speech perception, it is known that long reverberation causes degradation in speech comprehension. This is especially true for elderly people, the hearing-impaired, and non-native listeners. In order to prevent intelligibility degradation, we can apply a signal processing technique to the public address system of a room before speech signals are radiated through the loudspeakers. So far, we have developed several such preprocessing techniques. The two main techniques are modulation filtering [Kusumoto et al., Speech Commun. 45, 101–113 (2005)] and steady-state suppression [Arai et al., Acoust. Sci. Technol. 23, 229–232 (2002)]. Both techniques essentially enhance the temporal dynamics of speech between 1 and 16 Hz [Arai et al., J. Acoust. Soc. Am. 105, 2783–2791 (1999)]. Steady-state suppression, which suppresses steady-state portions of speech, reduces overlap-masking and improves speech intelligibility for young, elderly, and non-native listeners in reverberant environments [e.g., Hodoshima et al., J. Acoust. Soc. Am. 119, 4055–4064 (2006)]. This technique is more effective when speech-rate slowing is applied in advance [Arai et al., Acoust. Sci. Tech. 26, 459–461 (2005)]. Intelligibility with this technique exceeds a simple speech-rate slowing approach. [Work partially supported by JSPS.KAKENHI (16203041).]