The laryngeal control for pitch change has been studied in order to investigate the prosodic features of speech. This study used a set of models which simulate the characteristics of physiological control in the process of glottal sound generation from the neuromotor command. The advantage of such a method is that the control signals at various levels in the process can be examined by referring to the measured data, such as the electromyogram, the motion of larynx, and the acoustical nature of glottal sound. Electromygrams (EMG) were obtained from the utterance of Japanese word accentuation for five laryngeal muscles (Fig. 1). The speech samples were a nonsense Japanese mora sequence pronounced in isolation with four accentuation types in Tokyo dialect. The patterns of muscles activity, which we call the "continuous neuromotor commands", were obtained by the pulse counting of EMG signals (Fig. 2). The typical thyrogram, which indicates the vertical displacement of the larynx was also examined. As we have previously reported, the thyrogram shows good correspondence with the change in fundamental frequency. Five basic components were assumed to be involved in the continuous neuromotor command considering the linguistically significant characteristics of word accentuation in Japanese. The nature of these components was analyzed qualitatively by an analysis-by-synthesis method, and the components of higher command were derived from the continuous motor command. As the result, the components which specify Japanese word accentuation, and the one specifying the dialect characteristic, were observed separately in the higher command, and the distinction of the accented mora was made only by shifting the timing in the components of accentuation (Fig. 3). The cricothyroid muscle seems to be the primary control of pitch rise, as has been previous reported. In addition, existence of component, whose activity occurs during the decrease in pitch, suggests that pitch is lowered not only by a decrease of the cricothyroid muscle activity, but also by an increase of activity of some other muscles with it (Table 1). Although the involvement of extrinsic laryngeal muscles in the control of fundamental frequency seems to be secondary, the correlation between the activities of these muscles and the vertical displacement of the larynx during pitch change may be explained by reference to the schematized anatomical structure of the larynx (Fig. 4). Based on the mechanical model of the anatomical structure of the larynx (Fig. 5), a change in tension of each muscle and motion of the structure was simulated by the input of the higher commands (Fig. 8). Then, using a model of glottal sound generation (Fig. 6), changes in the fundamental frequency of the glottal sound and in the envelope of glottal volume velocity were derived (Fig. 8) by combining the change in the tension of vocal folds with the change in expiratory pressure through a functional relation of the vocal fold vibration (Fig. 7). Calculated patterns of the vertical displacement of the cricoid tip and change in fundamental frequency show a good correspondence to the measured counterparts (Fig. 9). The effect of the activity of extrinsic muscles on the change in fundamental frequency was also obtained using this model. The rate of the change in the fundamental frequency caused by the contraction of each muscle was estimated in this model by comparison with the change caused by the contraction of both the cricothyroid and vocalis muscles. The rate of increase of fundamental frequency caused by the contraction of the thyrohyoid muscle was estimated to be nearly the same as that caused by only the cricothyroid muscle, while the contraction of only the sternothyroid muscle caused some decrease of the fundamental frequency (Table 2). In this anatomical structure model, the hyoid bone was fixed and the simulation was done based on the

(View PDF for the rest of the abstract.)