Process for measuring the resemblance between sound samples and apparatus for performing this process
First Claim
1. A process for measuring a resemblance between a plurality of sound samples, comprising:
- a learning phase comprising the steps of;
acoustically analyzing a reference sound sample using a pth order acoustical analysis to obtain a resultant vector signal of the reference sound sample;
calculating a covariance matrix X of size p×
p using the resultant vector signal of the reference sound sample;
inverting said covariance matrix X to obtain an inverted covariance matrix X-1 ; and
storing the inverted covariance matrix X31 1 in a dictionary;
a test phase comprising the steps of;
acoustically analyzing a test sound sample using a pth order acoustical analysis to obtain a resultant vector signal of the test sound sample;
calculating a covariance matrix Y of size p×
p using the resultant vector signal of the test sound sample;
multiplying the covariance matrix Y with the inverted covariance matrix X-1 to obtain a product YX-1 ;
calculating p eigenvalues λ
k of YX-1 ;
determining a resemblance between the reference sound sample and the test sound sample using generalized sphericity functions using at least two of the following equations;
##EQU12## where a, g, and h represent respectively an arithmetic, geometric, and harmonic mean values of the eigenvalues λ
k,wherein the learning phase and test phase each further perform the steps of;
amplifying the sound samples;
filtering the amplified sound samples; and
digitizing the filtered amplified sound samples.
3 Assignments
0 Petitions
Accused Products
Abstract
A process and apparatus for measuring the resemblance between sound samples incorporating a learning phase and a test phase, each having a digital acquisition and preprocessing stage, an order p acoustic analysis stage and a p×p covariance matrix calculating stage with respect to the resulting vector signal, the learning phase also having a reference X covariance matrix inversion stage and a stage of storing the thus obtained matrix X-1 in a dictionary, the test phase incorporating a stage of multiplying the covariance matrix of test Y with the reference matrix X-1, a stage of extracting all or part of the p eigenvalues λk of said matrix product and a calculating stage using a family of functions f, called generalized sphericity functions, so as to obtain a digital value measuring the resemblance between the considered test sample and the reference samples of the dictionary. Alternative implementations are also described, which advantageously allow the same result to be obtained without requiring the explicit calculation of eigenvalues.
242 Citations
12 Claims
-
1. A process for measuring a resemblance between a plurality of sound samples, comprising:
-
a learning phase comprising the steps of; acoustically analyzing a reference sound sample using a pth order acoustical analysis to obtain a resultant vector signal of the reference sound sample; calculating a covariance matrix X of size p×
p using the resultant vector signal of the reference sound sample;inverting said covariance matrix X to obtain an inverted covariance matrix X-1 ; and storing the inverted covariance matrix X31 1 in a dictionary; a test phase comprising the steps of; acoustically analyzing a test sound sample using a pth order acoustical analysis to obtain a resultant vector signal of the test sound sample; calculating a covariance matrix Y of size p×
p using the resultant vector signal of the test sound sample;multiplying the covariance matrix Y with the inverted covariance matrix X-1 to obtain a product YX-1 ; calculating p eigenvalues λ
k of YX-1 ;determining a resemblance between the reference sound sample and the test sound sample using generalized sphericity functions using at least two of the following equations;
##EQU12## where a, g, and h represent respectively an arithmetic, geometric, and harmonic mean values of the eigenvalues λ
k,wherein the learning phase and test phase each further perform the steps of; amplifying the sound samples; filtering the amplified sound samples; and digitizing the filtered amplified sound samples. - View Dependent Claims (2, 3, 4, 5)
-
-
6. An apparatus for measuring a resemblance between a plurality of sound samples, comprising:
-
a learning means, including; means for acoustically analyzing a reference sound sample using a pth order acoustical analysis to obtain a resultant vector signal of the reference sound sample; means for calculating a covariance matrix X of size p×
p using the resultant vector signal of the reference sound sample;means for inverting said covariance matrix X to obtain an inverted covariance matrix X-1 ; and means for storing the inverted covariance matrix X-1 in a dictionary; a test means, including; means for acoustically analyzing a test sound sample using a pth order acoustical analysis to obtain a resultant vector signal of the test sound sample; means for calculating a covariance matrix Y of size p×
p using the resultant vector signal of the test sound sample;means for multiplying the covariance matrix Y with the inverted covariance matrix X-1 to obtain a product YX-1 ; means for calculating p eigenvalues λ
k of YX-1 ;means for determining a resemblance between the reference sound sample and the test sound sample using generalized sphericity functions using at least two of the following equations;
##EQU17## where a, g, and h represent respectively an arithmetic, geometric, and harmonic mean values of the eigenvalues λ
k,wherein the learning means and test means each include respective; means for amplifying the sound samples; means for filtering the amplified sound samples; and means for digitizing the filtered amplified sound samples. - View Dependent Claims (7, 8, 9)
-
-
10. An apparatus for measuring the resemblance between sound samples, comprising:
-
a digital acquisition and pre-processing module which converts an analog sound sample into a digital signal; an acoustic analysis module which transforms the digital signal into a vector signal composed of acoustic parameters; a covariance matrix module which computes a covariance of the acoustic vector signal at the output of the acoustic analysis module; a matrix inversion module for calculating the inverse of the symmetric matrix at the output of the covariance matrix module; a module for storing the output of the matrix inversion module in a dictionary, the storing module being only active during a training phase; a matrix multiplication module which takes during a test phase, the covariance matrix at the output of the covariance matrix module and multiplies it with an inverse covariance matrix from the dictionary, the matrix multiplication module being active only during the test phase; a module for extracting eigenvalues of the product obtained at the output of the matrix multiplication module, the extracting module being active only during the test phase; and a module for calculating a resemblance signal as a sphericity function of the eigenvalues produced by module using at least two of the following equations;
##EQU22## where a, g, and h represent respectively an arithmetic, geometrical, and harmonic mean values of the eigenvalues λ
k. - View Dependent Claims (11, 12)
-
Specification