This study conccrns the mathematical modelling of speech related acoustical hearing discomforts in medium to large environments (e.g. auditoriums and concert halls). Discomforts arise fro both phase distortion (echo) and perceived frequency transformations (Comb filters) Borwidk(1), Ando(3). These distortions are caused by the physical constraints imposed on sound transmission within large enviroments. Deconvolution was applied to the meeasured signals to recover the original sound from the environmentally distorted sound, thus scparating it from teh transformation system function(Room impulse response). Several samples were used for differnt positions wihtin the hall. The differnt methods investigated for the deconvolution process were cross-correlation, cepstrum and adaptive filter techniques, Oppcnhcim(2). Upon separation the inverse of the system transfer function was determined and used in the latter part of the study to pre-deform the original sound. Only a part of the system recsponsc was used. The first 20-30 ms constructively contribute to the intelligibility of speech and masked towards the end of the impulse response by the background noise at 40 dBA, Borwick(1). The critical part between 0.020 and 2 sec was then used as the basis of the filter lgortihm design. Because of the lengthof this part of the impulse response as well as the real time processing constraints a FIR filter could not be mplemented, Ando(3), Tohyama(4), Instead the filter was designed in the frequency domain. During transmission the environemtn now acting on the pre-deformed (filtered) sound renders beteter quality speech that is easier to understand, effectively removing part of the deformation. Finally the psychoacoustics of the total system were evaluated by a panel of listeners in the auditorium (concert hall) and the system was found to be effective for some targeted areas within the particular hall. The impact of these results for addressing acoustical problems will be discussed.
展开▼
