Speech recognition

• US 6,804,643 B1
• Filed: 10/27/2000
• Issued: 10/12/2004
• Est. Priority Date: 10/29/1999
• Status: Expired due to Fees

• Current Assignee: Nokia Mobile Phones UK Limited (Nokia Corporation)
• Original Assignee: Nokia Mobile Phones UK Limited (Nokia Corporation)
• Inventors: Kiss, Imre
• Primary Examiner(s): ABEBE, DANIEL DEMELASH
• Application Number:
  US09/698,805
• Time in Patent Office: 1,446 Days
• Field of Search: 704/243, 704/234, 704/235, 704/246, 704/261
• US Class Current: 704/234
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 25/27 characterised by the analysis technique

Abstract

A speech recognition feature extractor for extracting speech features from a speech signal, comprising: a time-to-frequency domain transformer (FFT) for generating spectral magnitude values in the frequency domain from the speech signal; a frequency domain filtering block (Mel) for generating a sub-band value relating to spectral magnitude values of a certain frequency sub-band; a compression block (LOG) for compressing said sub-band values; a transformation block (DCT) for obtaining a set of de-correlated features from the compressed sub-band values; and normalising block (CN) for normalising de-correlated features.

Claims

1. A speech recognition feature extractor for extracting speech features from a speech signal, comprising:

a time-to-frequency domain transformer for generating spectral magnitude values in the frequency domain from the speech signal;

a frequency domain filtering block for generating a sub-band value relating to spectral magnitude values of a certain frequency sub-band, for each of a group of frequency sub-bands;

a compression block for compressing said sub-band values;

a transformat on block for obtaining a set of de-correlated features from the sub-band values; and

a normalising block for normalizing features;

said feature extractor comprising a mean emphasising block for emphasizing at least one of the sub-band values after frequency domain filtering, wherein the emphasising is accomplished by addition of a mean value of sub-band signals to said at least one of the sub-band values.

2. A speech recognition feature extractor according to claim 1, wherein the mean emphasising block is arranged to mean emphasise all the sub-band values.

3. A speech recognition feature extractor according to claim 1, wherein the mean emphasising block is arranged to mean emphasise some of the sub-band values.

4. A speech recognition feature extractor according to claim 1, wherein said frequency domain filtering block is arranged to generate sub-band values according to a scale based on an auditory model.

5. A speech recognition feature extractor according to claim 1, comprising a differentiation block for generation of first time derivatives and second time derivatives for each of said de-correlated features;

and wherein

said normalising block is arranged to generate normalised speech features using said de-correlated features, said first derivatives features, and said second derivatives.

6. A speech recognition feature extractor according to claim 1, wherein, in said addition of the mean value, said mean emphasising block is arranged to add a mean estimate term to each sub-band value that is to be mean emphasised.

7. A speech recognition feature extractor according to claim 6, wherein the mean emphasising block is arranged to calculate the mean estimate term from compressed sub-band values representing a series of at least two subsequent speech frames.

8. A mobile station comprising a speech recognition feature extractor for extracting speech features from a speech signal, said extractor comprising:

a time-to-frequency domain transformer for generating spectral magnitude values in the frequency domain from the speech signal;

a frequency domain filtering block for generating a sub-band value relating to spectral magnitude values of a certain frequency sub-band, for each of a group of frequency sub-bands;

a compression block for compressing said sub-band values;

a transformation block for obtaining a set of de-correlated features from the sub-band values; and

a normalising block for normalising features;

said feature extractor comprising a mean emphasising block for emphasising;

at least one of the sub-band values after frequency domain filtering, wherein the emphasising is accomplished by addition of a mean value of sub-band signals to said at least one of the sub-band values.

9. A method for extracting speech features from a speech signal, comprising the steps of:

generating spectral magnitude values in the frequency domain from the speech signal;

generating a sub-band value relating to spectral magnitude, values of a certain frequency sub-band;

compressing said sub-band values;

obtaining a set of de-correlated features from the sub-band values;

normalising features; and

emphasising at least one of the sub-band values after frequency domain filtering, wherein the emphasising is accomplished by addition of a mean value of sub-band signals to said at least one of the sub-band values.

10. A computer program for extracting speech features from a speech signal, comprising:

a computer readable program means for causing a computer to generate spectral magnitude values in the frequency domain from the speech signal;

a computer readable program means for causing a computer to generate a sub-band value relating to spectral magnitude values of a certain frequency sub-band, for each of a group of frequency sub-bands;

a computer readable program means for causing a computer to compress said sub-band values;

a computer readable program means for causing a computer to obtain a skit of de-correlated features from the sub-band values;

a computer readable program means for causing a computer to normalise features; and

a computer readable mean-emphasising program means for causing a computer to emphasise at least one of the sub-band values after frequency domain filtering, wherein emphasising is accomplisher by addition of a mean value of sub-band signals to said at least one of the sub-band values.