High resolution post processing method for a speech decoder

• US 6,138,093 A
• Filed: 03/02/1998
• Issued: 10/24/2000
• Est. Priority Date: 03/03/1997
• Status: Expired due to Term

• Current Assignee: Telefonaktiebolaget LM Ericsson
• Original Assignee: Telefonaktiebolaget LM Ericsson
• Inventors: Kleijn, Bastiaan, Ekudden, Erik, Hagen, Roar
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Azad, Abul K.
• Application Number:
  US09/032,942
• Time in Patent Office: 967 Days
• Field of Search: 704/203, 704/205, 704/211, 704/226, 704/227, 704/228, 704/233, 704/278
• US Class Current: 704/228
• CPC Class Codes: G10L 19/26 Pre-filtering or post-filtering
• CPC Class Codes: G10L 21/0232 Processing in the frequency domain
• CPC Class Codes: G10L 25/27 characterised by the analysis technique

Abstract

A post-processing method for a speech decoder which outputs a decoded speech signal in the time domain provides high frequency resolution based on a frequency spectrum having non-harmonic and noise deficiencies. This is obtained by transforming the decoded time domain signal to a frequency domain signal by using a high frequency resolution transform (FFT). Then an analysis of the energy distribution of the frequency domain signal is made throughout its frequency area (4 kHz) to find the disturbing frequency components and to prioritize such frequency components which are situated in the higher part of the frequency spectrum. Next, the suppression degree for the disturbing frequency components is found based on prioritizing. Finally the steps of controlling a post-filtering of the transform in dependence of the finding, and inverse transforming the post-filtered transform in order to obtain a post-filtered decoded speech signal in the time domain are performed.

Claims

1. A method for post-processing a decoded time domain signal received from a speech decoder in order to reduce non-harmonic and noise deficiencies within said signal, said method comprising the steps of:

a) performing a high-frequency resolution transform on the decoded signal to obtain a frequency spectrum of the decoded speech signal;

b) analyzing said frequency spectrum by estimating likely coding noise characteristics in various frequency areas based on the properties of the coding algorithm of the decoder from which the decoded signal was received, to identify disturbing frequency components;

c) identifying a degree of suppression for the disturbing frequency components; and

d) performing high frequency resolution filtering of said frequency spectrum in order to significantly reduce disturbing frequency components in said frequency areas, based on the degree of suppression for the disturbing frequency components found in step c.

2. The method in claim 1, wherein said step of analyzing said frequency spectrum in various frequency areas is further based on decoder attributes.

3. The method in claim 1, wherein said step of analyzing said frequency spectrum in various frequency areas is further based on a perceptual model.

4. The method in claim 1, wherein said high frequency resolution filtering is further based on dynamic properties of the filter.

5. The method in claim 4, wherein said high frequency resolution filtering is further based on dynamic properties of the decoded signal.

6. A method for post-processing a decoded time domain signal received from a speech decoder in order to reduce non-harmonic and noise deficiencies in said signal, said method comprising the steps of:

a) transforming the decoded time domain signal to a frequency domain signal by means of a high frequency resolution transform (FFT);

b) analyzing the energy distribution of said frequency domain signal throughout its frequency area to find disturbing frequency components and to prioritize said disturbing frequency components which are situated in the higher part of the frequency spectrum;

c) finding a degree of suppression for said disturbing frequency components based on the prioritization of said disturbing frequency components;

d) post-filtering said frequency domain signal in dependence of the degree of suppression found in step c; and

e) inverse transforming the post-filtered frequency domain signal in order to obtain a post-filtered decoded speech signal in the time domain.

7. Method according to claim 6, wherein said step of analyzing the energy distribution of said frequency domain signal comprises:

a) detecting the envelope of a signal representing said frequency spectrum and forming a corresponding envelope signal;

b) estimating the slope of said signal representing the frequency spectrum and forming a corresponding slope signal; and

wherein said step of post-filtering said frequency domain signal comprises the steps of;

c) comparing said signal representing the frequency spectrum with said slope signal in order to locate said disturbing frequency components;

d) forming a value representing a degree of suppression for a specific frequency component based on the result of said comparing and said signal corresponding to the slope; and

e) repeating said step of forming a value representing the degree to suppress a specific frequency component in order to obtain a number of values, said values being used to control said post-filtering of the frequency domain signal.

8. Method according to claim 6, further comprising the step of:

smoothing the frequency domain signal.