Abstract:
In this paper, a new method to calculate Japanese CV monosyllable speech intelligibility (confusion rate) by use of an autocorrelation function (ACF) is proposed. This method is based on the principle of speech information processing in the auditory pathway and the brain. In order to find out the essential part to identify syllables, the head and tail parts of syllables were cut and speech intelligibility tests were conducted. The normalized ACF of the essential part was analyzed. Results show that all syllables may be characterized by the time ((tau)) and the amplitude ((phi)) of only the first dip of the normalized ACF. Using this characterization, together with the distance between syllables and the manner of articulation, the monosyllable speech confusion rate may be calculated.