As a part of development program for the diagnostic system of Japanese misarticulation, the speech analysing system using a microcomputer (NEC PC 9801 VM 21) combined with high-speed palatography was developed, and distorted sounds with Japanese Palatalized Articulation (JPA), often associated with cleft palate speech were analysed.
Twenty-eight cleft palate patients, and one functional misarticulation patient suffering from JPA with sufficient velopharyngeal function, and as a control, thirty-one persons with normal speech served for this study.
Target syllables were normal and JPA of /ta/, /da/,/tsw/,/dzw/ and /sa/, which are judged as JPA in high frequency, and /ka/, /ga/, /kw/ and /gw/ which were often confused with JPA in clinical examination.
Acoustical analysis were carried out by frequency and time domain for these target syllables and the following results were obtained.
1) On the consonant spectrum envelope (SE) normalized for level, the differences between high and low subband level were calculated and defined as Spectrum Envelope Score (SES). As the result, although normal /sa/ and it's JPA could be distinguished, other normal and JPA could not.
2) The maximum level of frequency on the consonant SE was defined as Consonant Peak Energy Frequency (CPF). JPA of /ta/,/da/,/tw/ and /dzw/ were distinguished from normal articulation of them, and /ka/, /da/, /kw/ and /gw/ respectively by CPF.
3) The second and the third formant variance between transient portion was calculated by LPC method (ΔF2, ΔF3). Distinction of normal /ta/, /da/, /tsw/ and /dzw/ and it's JPA were possible by calculation of ΔF2-ΔF3 respectively. However, JPA and it's corresponding velar sound /ka/, /ga/, /kw/ and /gw/ was not distinguished.
4) As for VOT, JPA of /ta/ and /da/ had similar scores to velar sound /ka/ and /ga/respectively different from that of normal sound. But, it was unable to distinctly determine normal and JPA by VOT.
5) As CPF and ΔF2-ΔF3 were considered to be suitable for the quantitative evaluation of JPA from the above mentioned results, multi revolution analysis were carried out on them and the estimated scores from psycho-acoustical judgement. Although a close correlation between the physical judgment score and clinical confusion tendency was proved, correlation between the score and the characteristic noise caused by abnormal tongue movement was not found.
View full abstract