GB1303808A - - Google Patents

Abstract

1303808 Formant tracking vocoder PLESSEY CO Ltd 22 April 1970 [24 April 1969] 21000/69 Heading H4R In a formant tracking vocoder, Fig. 1, speech SP is fed to two channels to derive two formant frequencies of speech. Each channel includes a band stop filter F -1 or F -2 tuned to reject one formant frequency and a frequency discriminator F 2T or FIT to detect the frequency of the other formant. The output of the discriminators in each channel are utilized to tune the band stop filter in the other channel to reject the appropriate formant. A third channel may be provided including two band stop filters F -11 and F -22 to reject the two formants, each being tuned by the appropriate discriminator output, in order to provide a signal in which the pitch frequency can easily be determined by discriminator C 0 . Each formant discriminator may include a band reject filter with a phase sensitive detector connected between input and output with the integrated output of the detector controlling the filter tuning, and providing the discriminator output. In this arrangement the output from the band reject filter is a suitable signal from which the pitch frequency can be determined. Three formant tracker, Fig. 4.-The channel which includes the filter F -2 to reject the middle formant is fed to low pass and high pass filters LP1 and HP1 in order to separate out the lowest and highest frequency formants which are detected by respective discriminators. The outputs from these discriminators control respective filters F -1 and F -3 in the other channel in order to detect the middle formant. The three discriminator outputs control band stop filters F -11 , F -22 and F -33 in the pitch discriminator channel.