Complex cepstrum analyzer for speech signals

• US 5,677,984 A
• Filed: 02/23/1995
• Issued: 10/14/1997
• Est. Priority Date: 02/23/1994
• Status: Expired due to Term

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Mitome, Yukio
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Mattson, Robert
• Application Number:
  US08/392,482
• Time in Patent Office: 964 Days
• Field of Search: 395/2.12, 395/2.13, 395/2.14, 395/2.2, 395/2.4, 381/29, 381/30, 381/37, 381/40
• US Class Current: 704/220
• CPC Class Codes: G01R 23/16 Spectrum analysis; Fourier analysis
• CPC Class Codes: G10L 25/24 the extracted parameters being the cepstrum
• CPC Class Codes: G10L 25/48 specially adapted for particular use

Abstract

Respective samples of input speech data are transformed by a discrete Fourier transformer to obtain a spectrum of the speech data. Simultaneously, values of times of the respective samples are multiplied with the input speech data by a multiplier and a differential spectrum is obtained by transforming a result of the multiplication by a discrete Fourier transformer. A real part of a value obtained by dividing the differential spectrum by the spectrum by a quotient real part calculator and the real part is inverse-transformed by a discrete inverse Fourier transformer. The result of the inverse-transformation is divided by the values of the times of the respective samples to obtain a time function corresponding to phase. On the other hand, a time function corresponding to a logarithmic amplitude spectrum is obtained from an output of the inverse-Fourier transformer by means of a logarithmic amplitude spectrum calculator. A complex cepstrum is obtained by adding the both time functions at respective times by an adder.

Claims

1. A complex cepstrum analyzer for a speech signal, for analyzing the speech signal and extracting a complex cepstrum, comprising:

a first discrete Fourier transformer for performing a discrete Fourier transformation of samples x(n) of an input speech signal and outputting a spectrum of the input speech signal as a voice spectrum, wherein n denotes a variable indicative of a sample time of each of the samples;

a multiplier for multiplying the samples of the speech signal with values of the sample time n corresponding to each of the samples;

a second discrete Fourier transformer for performing a discrete Fourier transformation for an output of said multiplier and outputting a result as a differential spectrum, the result being a differentiation of said voice spectrum multiplied with a unit imaginary number;

a logarithmic amplitude spectrum calculator for calculating a logarithm of an absolute value of the voice spectrum output from said first discrete Fourier transformer and outputting a result as a logarithmic amplitude spectrum;

a first discrete inverse Fourier transformer for performing a discrete inverse Fourier transformation for the logarithmic amplitude spectrum outputted from said logarithmic amplitude spectrum calculator and outputting a result as a time function corresponding to the logarithmic amplitude spectrum;

a quotient real part calculator for outputting a real part of a quotient obtained by dividing the differential spectrum output from said second discrete Fourier transformer by the voice spectrum output from said first discrete Fourier transformer;

a second discrete inverse Fourier transformer for performing a discrete inverse Fourier transformation for the real part of the quotient output from said quotient real part calculator and outputting a result as a time function value corresponding to the real part of the quotient;

a divider for dividing a value of each sample of the time function corresponding to the real part of the quotient by a value of each time and outputting a result as a time function value corresponding to phase of the spectrum; and

an adder for adding time values of the time function corresponding to phase of the voice spectrum output from said divider and values of the time function corresponding to the logarithmic amplitude spectrum for each sample time n and outputting a result as values of said complex cepstrum for each sample time n.

2. A complex cepstrum analyzer for voice signals, for analyzing an input speech signal and extracting a complex spectrum, comprising:

a first buffer connected to an input terminal for receiving samples x(n) of an input speech signal, wherein n denotes a variable indicative of a sample time of each of the samples;

a multiplier connected to said input terminal, for multiplying the samples x(n) of the input speech signal with values of the sample time n corresponding to each of the samples;

a second buffer connected to said multiplier and receiving an output of said multiplier;

a discrete Fourier transformer connected to said first and second buffers for performing a discrete forward or inverse Fourier transformation of an output received from first buffer and an output received from said second buffer and sending a result back to one of said first and second buffers to which the output was received, said first buffer being adapted to store a spectrum obtained from said discrete Fourier transformer by said discrete Fourier transformer performing a discrete Fourier transformation for the samples of the input speech signal data stored in said first buffer, and said second buffer being adapted to store a differential spectrum obtained from said discrete Fourier transformer by said discrete Fourier transformer performing a discrete Fourier transformation for the samples of the data output from said multiplier and stored in said second buffer;

a quotient real part calculator for sending a real part of a quotient obtained by dividing the differential spectrum output from said second buffer by the spectrum output from said first buffer back to said second buffer;

a divider for dividing a value of each sample of a time function corresponding to the result of an inverse Fourier transformation performed by said discrete Fourier transformer and sent back to said second buffer and stored in said second buffer by reference to the sample time n, and sending a result back to said second buffer as a time function corresponding to a phase of the spectrum;

a logarithmic amplitude spectrum calculator for calculating a logarithm of an absolute value of the spectrum stored in said first buffer and sending a result back to said first buffer as a logarithmic amplitude spectrum, the logarithmic amplitude spectrum being then sent from said first buffer to said discrete Fourier transformer to be subject to an inverse Fourier transformation by said discrete Fourier transformer to obtain a time function of the logarithmic amplitude spectrum which is stored in said first buffer; and

an adder for adding time values of the time function corresponding to the logarithmic amplitude spectrum stored in said first buffer and the time function corresponding to the phase of the spectrum stored in said second buffer at each sample time n, to obtain values of a complex cepstrum at each sample time n and outputting the values of the complex cepstrum at each sample time n.

3. A complex cepstrum analyzer for a speech signal, for analyzing the speech signal and extracting a complex cepstrum, comprising:

a first buffer connected to an input terminal for receiving an input speech signal;

a multiplier connected to said input terminal, for multiplying samples x(n) of the input speech signal with values n representing sampling times of the samples;

a second buffer connected to said multiplier for temporarily storing an output of said multiplier;

a discrete Fourier transformer connected to said first and second buffers for performing a discrete forward or inverse Fourier transformation of an output received from first buffer and an output received from said second buffer and sending a result back to one of said first and second buffers to which the output was received, said first buffer being adapted to store a spectrum obtained from said discrete Fourier transformer by said discrete Fourier transformer performing a discrete Fourier transformation for the samples of the input speech signal data stored in said first buffer, and said second buffer being adapted to store a differential spectrum obtained from said discrete Fourier transformer by said discrete Fourier transformer performing a discrete Fourier transformation for the samples of the data output from said multiplier and stored in said second buffer;

a quotient real part calculator for sending a real part of a quotient obtained by dividing the differential spectrum output from said second buffer by the spectrum output from said first buffer back to said second buffer;

a divider for dividing a result of inverse transformation obtained by an inverse transformation of the real part of the quotient stored in said second buffer by means of said discrete Fourier transformer, sent back to said second buffer and stored in said second buffer by the time values n corresponding to the time samples of the time function, and sending a result back to said second buffer as a time function corresponding to a phase of the voice signal spectrum;

an argument calculator for obtaining, after said discrete Fourier transformer receives the time function corresponding to the phase of the voice spectrum from said second buffer and performs a discrete Fourier transformation for the time function to obtain the phase of the spectrum and sends the phase of the spectrum back to said second buffer, an argument by calculating arc tangent of a real part and an imaginary part of the spectrum supplied from said first buffer;

a phase corrector for receiving the phase of the spectrum stored in said second buffer, correcting the phase of the spectrum with a value of argument calculated by said argument calculator, and outputting the corrected phase to said second buffer;

a logarithmic amplitude spectrum calculator for calculating, after said discrete Fourier transformer performs a discrete Fourier transformation for the corrected phase of the spectrum stored in said second buffer to obtain the time function corresponding to the corrected phase of the spectrum and sends the time function corresponding to the corrected phase back to said second buffer, a logarithm of an absolute value of the spectrum stored in said first buffer and sending a result back to said first buffer as a logarithmic amplitude spectrum, the logarithmic amplitude spectrum being then sent from said first buffer to said discrete Fourier transformer to be subject to an inverse Fourier transformation by said discrete Fourier transformer to obtain a time function of the logarithmic amplitude spectrum which is stored in said first buffer; and

an adder for adding time values of the time function corresponding to the logarithmic amplitude spectrum obtained by said discrete Fourier transformer performing the inverse transformation for the logarithmic amplitude spectrum stored in said first buffer by means of said discrete Fourier transformer and sent back to and stored in said first buffer and the time function corresponding to the phase of the voice signal spectrum corrected and stored in said second buffer at each sample time n, to obtain values of a complex cepstrum at each sample time n and outputting a result to an output terminal.