Real time cepstrum analyzer

Claim 2

Apparatus for continuously determining whether a speech wave is periodic or aperiodic, which comprises: means supplied with a time varying speech wave for selecting therefrom successive wave segments, each of which contains at least two periods of said speech wave; first spectrum analyzer means supplied with said selected speech wave segments for developing short-time spectrum samples of each of said speech wave segments; logarithmic amplifier means supplied with said spectrum samples for developing a signal proportional to the logarithm of said samples; second spectrum analyzer means supplied with said logarithmic signals for developing short-time spectrum samples thereof; and means, responsive tO said spectrum samples produced by said second analyzer means, for indicating the presence or absence of a single large peak among said samples, the presence of a single large peak designating said speech wave as periodic, and the absence of a large peak designating said speech wave as aperiodic.

Claim 3

Apparatus as defined in claim 1 wherein said means for selecting successive wave segments from said supplied time varying speech wave is adjusted to select speech wave segments which overlap one another by a predetermined amount.

Claim 4

Apparatus as defined in claim 1, in combination with means, supplied with short-time spectrum samples produced by said second analyzer, for indicating the relative location of each indicated peak on a time scale as an indication of the pitch of said supplied speech wave.

Claim 5

Apparatus for continuously determining in real time whether a relatively long speech wave is periodic or aperiodic, which comprises: time compressor means supplied with a time varying speech wave for compressing the time scale of said speech wave; means for separating said compressed speech wave into segments of sufficient duration to contain at least two periods of said speech wave during periodic portions thereof; first spectrum analyzer means supplied with said compressed speech wave segments for developing short-time spectrum samples of each of said speech wave segments; logarithmic amplifier means supplied with said spectrum samples for developing a signal proportional to the logarithm of said samples; second time compressor means supplied with said logarithmic signals for compressing the time scale thereof; second spectrum analyzer means supplied with said compressed logarithmic signals for developing short-time spectrum samples thereof; and means, responsive to said spectrum samples produced by said second analyzer means, for indicating the presence of large peaks among said samples, the presence of a single large peak designating said speech wave as periodic and the absence of a single large peak designating said speech wave as aperiodic.

Claim 6

Apparatus for continuously determining in real time whether a relatively long speech wave is periodic or aperiodic, which comprises: time compressor means supplied with a time varying speech wave for compressing the time scale of said speech wave; means for separating said compressed speech wave into segments of sufficient duration to contain at least two periods of said speech wave during periodic portions thereof; first spectrum analyzer means supplied with said compressed speech wave segments for developing short-time spectrum samples of each of said speech wave segments; logarithmic amplifier means supplied with said spectrum samples for developing a signal proportional to the logarithm of said samples; second time compressor means supplied with said logarithmic signals for compressing the time scale thereof; second spectrum analyzer means supplied with said compressed logarithmic signals for developing short-time spectrum samples thereof; and means, responsive to said spectrum samples produced by said second analyzer means, for indicating the presence of large peaks among said samples, the presence of a single large peak designating said speech wave as periodic and the absence of a single large peak designating said speech wave as aperiodic.

Claim 7

Apparatus as defined in claim 4, in combination with means, responsive to said spectrum samples produced by said second analyzer means, for indicating the relative location of each indicated large peak among said samples on a time scale as an indication of the pitch of said supplied speech wave.

Claim 8

Apparatus for determining the pitch characteristic of a speech wave characterized by periodic and aperiodic portions, which comprises: means for dividing a supplied speech wave into successive overlapping segments, each having a predetermined length sufficient to contain at least two of the longest periods of said speech wave that are present during periodic portions thereof; means supplied with said successive speech wave segments for deriving a succession of cepstrum signals for each corresponding succession of said overlapping segments of said speech wave, each of said cepstrum signals indicating periodicity or aperiodicity in the corresponding speech wave segment by the presence or absence of a single large peak; and peak detector means for deriving from said succession of cepstrum signals a pitch control signal indicative of the presence or absence of a single large peak in successive cepstrum signals and, if a peak is present, for indicating the relative location of said peak on a time scale as an indication of the pitch of said corresponding speech wave segment.

Claim 9

Apparatus for determining the pitch characteristic of a speech wave characterized by periodic and aperiodic portions, which comprises: means for dividing a supplied speech wave into successive overlapping segments, each having a predetermined length sufficient to contain at least two of the longest periods of said speech wave that are present during periodic portions thereof; means supplied with said successive speech wave segments for deriving a succession of cepstrum signals for each corresponding succession of said overlapping segments of said speech wave, each of said cepstrum signals indicating periodicity or aperiodicity in the corresponding speech wave segment by the presence or absence of a single large peak; and peak detector means for deriving from said succession of cepstrum signals a pitch control signal indicative of the presence or absence of a single large peak in successive cepstrum signals and, if a peak is present, for indicating the relative location of said peak on a time scale as an indication of the pitch of said corresponding speech wave segment.

Claim 10

Apparatus for determining the pitch characteristic of an applied speech wave characterized by periodic portions, representing voiced sounds, and aperiodic portions, representing unvoiced sounds, which comprises: means for dividing an applied speech wave into successive segments, each segment having a predetermined uniform length containing at least two periods of said speech wave; first analyzer means for deriving from each of said segments a first plurality of samples representing selected values of a first short-time spectrum of said segment; means for converting said first plurality of samples into a logarithm waveform representing the logarithm of said first short-time spectrum; a second analyzer means supplied with said logarithm waveform for deriving therefrom a second plurality of samples representing selected values of a short-time spectrum of the logarithm of said first short-time spectrum; and means responsive to signals from said second spectrum analyzer for detecting the presence of a single large peak as an indication of a periodic segment of said applied speech wave.

Claim 11

Apparatus for determining the pitch characteristic of an applied speech wave characterized by periodic portions, representing voiced sounds, and aperiodic portions, representing unvoiced sounds, which comprises: means for dividing an applied speech wave into successive segments, each segment having a predetermined uniform length containing at least two periods of said speech wave; first analyzer means for deriving from each of said segments a first plurality of samples representing selected values of a first short-time spectrum of said segment; means for converting said first plurality of samples into a logarithm waveform representing the logarithm of said first short-time spectrum; a second analyzer means supplied with said logarithm waveform for deriving therefrom a second plurality of samples representing selected values of a short-time spectrum of the logarithm of said first short-time spectrum; and means responsive to signals from said second spectrum analyzer for detecting the presence of a single large peak as an indication of a periodic segment of said applied speech wave.

Claim 12

In a vocoder communication system, apparatus for deriving a pitch control signal from an applied speech wave, which comprises: means for dividing an applied speech wave into successive segments, each segment having a predetermined uniform length sufficient to contain at least two of the longest periods of said speech wave that are present during periodic portions thereof; means, supplied with successive segments of said speech wave, for deriving signals representing short-time power spectra of said speech wave segments; means for developing, from said short-time spectra signals, signal waves that represent the logarithms of said short-time power spectra; means supplied with said logarithm signals for deriving signals representing the short-time power spectra of said logarithm signals; means for detecting the presence of a single large peak in each of said signals representing the spectrum of one of said logarithm signals as an indication of a periodic segment of said speech wave; and means for indicating the relative location of each such detected peak on a time scale as an indication of the momentary pitch of said speech wave.

Claim 13

In a vocoder communication system, apparatus for deriving a pitch control signal from an applied speech wave, which comprises: means for dividing an applied speech wave into successive segments, each segment having a predetermined uniform length sufficient to contain at least two of the longest periods of said speech wave that are present during periodic portions thereof; means, supplied with successive segments of said speech wave, for deriving signals representing short-time power spectra of said speech wave segments; means for developing, from said short-time spectra signals, signal waves that represent the logarithms of said short-time power spectra; means supplied with said logarithm signals for deriving signals representing the short-time power spectra of said logarithm signals; means for detecting the presence of a single large peak in each of said signals representing the spectrum of one of said logarithm signals as an indication of a periodic segment of said speech wave; and means for indicating the relative location of each such detected peak on a time scale as an indication of the momentary pitch of said speech wave.