Feature extraction method for a speech signal

• US 5,732,388 A
• Filed: 01/11/1996
• Issued: 03/24/1998
• Est. Priority Date: 01/10/1995
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Siemens AG
• Inventors: Hauenstein, Alfred, Marschall, Erwin, Hoege, Harald
• Primary Examiner(s): Tung, Kee M.
• Application Number:
  US08/584,816
• Time in Patent Office: 803 Days
• Field of Search: 395/2.1, 395/2.12, 395/2.14, 395/2.15, 395/2.3, 395/2.35-2.37, 395/2.39, 395/2.4, 395/2.42, 395/2.45, 395/2.6, 395/2.65
• US Class Current: 704/205
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 25/24 the extracted parameters being the cepstrum

Abstract

The invention expands a feature extraction unit of a type known in the art with an adaptive channel compensation. The filter coefficients for a high-pass filter for eliminating the chronologically sliding average are set time-dependently with the adaptive channel compensation. As a result, a faster channel adaptation, and a better error compensation in the speech feature extraction are achieved, leading to a more reliable speech recognition. The parameters for the individual function blocks of the feature extraction unit are determined and stored in advance for specific speech transmission channels such as in mobile radio telephone, ISDN and analog speech transmission. The speech recognition with the assistance of the feature extraction unit can subsequently then ensue in channel-associated fashion with the stored parameters values.

Claims

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

determining spectral energies (x₁, . . . , x_N) of a speech signal by brief duration spectral analysis;

obtaining logarithmized spectral energies (y₁, . . . , y_n) from said spectral energies (x₁, . . . x_N); and

space="preserve" listing-type="equation">y.sub.i (n)=y.sub.i '"'"'(n) -y.sub.i +(n) with i=1, . . . , K

forming channel-compensated spectral energies (y₁, . . . y_N) from individual spectral energies y_i (n) using a high-pass filter with time-dependent filter coefficients, wherein ##EQU8## wherein y_i0, α

_i0 and β

₀ and are constants that are acquired from statistical examinations of the speech signal, and

n(n) is a function monotonously rising with n.

2. A method as claimed in claim 1 comprising the additional steps of:

superimposing spectral energies of at least one known noise signal on the speech signal and determining noise energies (e_i); and

conducting an irrelevancy reduction of the speech signal by comparing at least one spectral energy (x_i) and a noise energy (e_i) from the same frequency band end, if the noise energy (e_i) is higher that the at least one spectral energy (x_i), using the noise energy (e_i) for further-processing of the speech signal instead of the spectral energy (x_i).

3. A method as claimed in claim 2, wherein the known noise signal is the quantization noise of a speech coder.

4. A method as claimed in claim 2, comprising the additional step of using frequencies that are located outside of a range defined for the speech signal as said known noise signal.

5. A method as claimed in claim 2, comprising the additional step of noise signal of background noises during speaking as said known noise signal.

6. A method as claimed in claim 1 comprising the additional steps of:

calculating and storing parameters which are characteristic for at least two speech transmission channels on which the speech signal is transmitted;

assigning an unambiguous identifier to each of the speech transmission channels; and

forming said channel-compensated spectral energies (y₁, . . . y_N) with the stored parameters dependent on the speech transmission channel on which said speech signal is transmitted.

7. A method as claimed in claim 6, wherein the step of calculating and storing said parameters comprises calculating and storing parameters for ISDN speech transmission.

8. A method as claimed in claim 6, wherein the step of calculating and storing said parameters comprises calculating and storing parameters for mobile radio telephone speech transmission.

9. A method as claimed in claim 6, wherein the step of calculating and storing said parameters comprises calculating and storing parameters for analog telephone speech transmission.