Variable-subframe-length speech-coding classes derived from wavelet-transform parameters

• US 5,781,881 A
• Filed: 10/21/1996
• Issued: 07/14/1998
• Est. Priority Date: 10/19/1995
• Status: Expired due to Term

• Current Assignee: Deutsche Telekom AG
• Original Assignee: Deutsche Telekom AG
• Inventors: Stegmann, Joachim
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US08/734,657
• Time in Patent Office: 631 Days
• Field of Search: 704/211, 704/214
• US Class Current: 704/211
• CPC Class Codes: G10L 19/18 Vocoders using multiple modes
• CPC Class Codes: G10L 2025/786 Adaptive threshold
• CPC Class Codes: G10L 25/27 characterised by the analysis technique
• CPC Class Codes: G10L 25/93 Discriminating between voiced and unvoiced parts of speech signals G10L25/90 takes precedence

Abstract

A method and a device are described for classifying speech on the basis of the wavelet transformation for low-bit-rate speech coding processes. The method and the device permit a more robust classifier of speech signals for signal-matched control of speech coding processes in order to reduce the bit rate without affecting the speech quality or to increase the quality at the same bit rate. The method provides that, after segmenting the speech signal, a wavelet transformation is calculated for each frame, from which a set of parameters is determined with the help of adaptive thresholds. The parameters control a finite-state model, which subdivides the frames into shorter subframes if required, and classifies each subframe into one of several classes typical for speech coding. The speech signal is classified on the basis of the wavelet transformation for each time frame. Thus both a high time resolution (location of pulses) and frequency resolution (good mean values) can be achieved. This method and the classifier are therefore especially well suited for the control and selection of code books in a low-bit-rate speech coder. They also have a low sensitivity to background noise and low complexity.

Claims

1. A method for classifying speech signals comprising the steps of:

segmenting the speech signal into frames;

calculating a wavelet transformation;

obtaining a set of parameters (P₁ -P₃) from the wavelet transformation;

dividing the frames into subframes using a finite-state model which is a function of the set of parameters;

classifying each of the subframes into one of a plurality of speech coding classes.

2. The method as recited in claim 1 wherein the speech signal is segmented into constant-length frames.

3. The method as recited in claim 1 wherein at least one frame is mirrored at its boundaries.

4. The method as recited in claim 1 wherein the wavelet transformation is calculated in smaller intervals, and the frame is shifted by a constant offset.

5. The method as recited in claim 1 wherein an edge of at least one frame is filled with previous or future sampling values.

6. The method as recited in claim 1 wherein for a certain frame s(k), a time-discrete wavelet transformation S_h (m,n) is calculated in reference to a certain wavelet h(k) with integer scaling (m) and time shift (n) parameters.

7. The method as recited in claim 6 wherein the set of parameters are scaling difference (P₁), time difference (P₂), and periodicity (P₃) parameters.

8. The method as recited in claim 7 wherein the set of parameters are determined from the transformation coefficients of S_h (m, n).

9. The method as recited in claim 1 wherein the set of parameters is obtained with the help of adaptive thresholds, threshold values required for obtaining the set of parameters being adaptively controlled according to a current level of background noise.

10. A method for classifying speech signals comprising the steps of:

segmenting the speech signal into frames;

calculating a wavelet transformation;

obtaining a set of parameters (P₁ -P₃) from the wavelet transformation;

dividing the frames into subframes based on the set of parameters, so that the subframes are classified as either voiceless, voicing onsets, or voiced.

11. A speech classifier comprising:

a segmentator for segmenting input speech to produce frames;

a wavelet processor for calculating a discrete wavelet transformation for each segment and determining a set of parameters (P₁ -P₃) with the help of adaptive thresholds; and

a finite-state model processor, which receives the set of parameters as inputs and in turn divides the speech frames into subframes and classifies each of these subframes into one of a plurality of speech coding classes.