| 摘要 |
1,225,142. Speech synthesizer. HITACHI Ltd. 27 Nov., 1968 [29 Nov., 1967], No. 56386/68. Heading H4R. In a speech synthesizer voiced sounds are reproduced by reading out from a store, and adding together a number of damped sinusoids, the period between successive read-outs being varied by a control signal, while consonants, i.e. unvoiced and nasal sounds, are produced from pre-recorded consonants under the control of further control signals. As described for voiced sounds three formants are generated separately and combined, each being synthesized by reading out from a magnetic recording on a drum one of a number of damped oscillations recorded on respective tracks on the drum. Each track has a number, e.g. 10, heads spaced at equal distances around the drum periphery and the heads are switched in sequence so as to reproduce the recorded damped oscillation at the desired repetition frequency, i.e. the pitch frequency of the speech to be synthesized. Each track has a damped oscillation of different frequency, there being, for example, 16 tracks available to synthesize the first formant with damped oscillations on adjacent tracks spaced by 50 Hz over the band 200 to 950 Hz. Similar arrangements provide the second and third formants. Fig. 5 shows an embodiment of a synthesizer for producing a number of speech signals on respective lines 31-3n and comprising a high-speed computer 10 which produces control signals on lines 11 to In, with reference to a pronounce dictionary in a store 20, to control the synthesis of a required sentence. The control signals are decoded in decoders 101 to 10n to produce signals at the appropriate times to select the frequencies of the formants and to switch the heads of the respective recorded tracks on drum 30 to reproduce the damped oscillations thereon at the required repetition frequency. The levels of the three formants in each speech channel are controlled relatively by modulators 311 and 312, for the top channel, before the three formants are combined when the level of the resultant signal is regulated in modulator 315 before the formants are added to the consonants, selected from further tracks on drum 30, and level regulated in modulator 313, to appear on output terminal 31. A suitable matrix of gates for selecting the appropriate reading heads is described with respect to Fig. 7 (not shown). |
