平原達也
時空間的なマスキングパターンから見た聴覚系内におけ
る音声スペクトル表現
Abstract:In this paper, several speech sounds are examined by a masking method to show typical
examples of speech spectrum in the auditory pathway represented by a spatio-temporal
masking pattern and to clarify differences between interaural and physical representation of
speech spectrum. Three types of Japanese speech, monosyllables, continuous speech and a
monosyllable reproduced time reversely, are chosen for masker sounds. Using 1/3 octave
band noise bursts with 25msec. duration as maskees, simultaneous and temporal masking
are measured for the whole period of each masker. Spatio-temporal masking patterns thus
obtained are an inter-aural speech spectrum. Compared with the physical spectral pattern: speech onsets and the formant structure, in particular, the transition of formants are emphasized and represented prominent in the masking patterns. These spectral emphases in the auditory pathway are composed of three functions, AM/FM masking, forward/backward masking, and adaptation. Further, taking into account the considerable differences between inter-aural and physical representation of speech spectrum, the inter-aural spectrum can be implemented as better representation of speech spectrum in speech feature extraction and speech signal processing by computers.