Extension of the bandwidth of the JSRU parallel-formant synthesizer for high quality synthesis of male and female speech
Wendy J. Holmes, John N. Holmes, M. W. Judd · International Conference on Acoustics, Speech, and Signal Processing · 2002
Extensions to the joint speech research unit (JSRU) parallel-formant synthesizer to enable synthesis in the frequency range up to 8 kHz instead of the 4 kHz used in the original version are described. The previous fixed F4 filter has been replaced by four fixed bandpass filters to completely cover the frequency range from just above 3 kHz up to 8 kHz. These filters have center frequencies of 3500 Hz, 4350 Hz, 5400 Hz, and 7000 Hz with progressively increasing bandwidths. Fixed filters are considered to be sufficient for the F4 region and above as it has been demonstrated that, as long as the overall signal level at high frequencies is appropriate, the fine detail is not important. The largely automatic copy synthesis procedure developed for use with the original JSRU synthesizer is modified for use with the extended version. Amplitudes are measured at the linear predictive coding derived formant frequencies from a fast Fourier transform analysis and then transformed using a table of amplitude correction values. High-quality copy synthesis is obtained for both male and female speech, with the nature and extent of the improvement dependent on the speaker characteristics.>