Methods and devices for noise conditioning signals representative of audio information in compressed and digitized form

• US 6,026,356 A
• Filed: 07/03/1997
• Issued: 02/15/2000
• Est. Priority Date: 07/03/1997
• Status: Expired due to Term

• Current Assignee: Genband US LLC (Ribbon Communications, Inc.)
• Original Assignee: Nortel Networks Corporation
• Inventors: Yue, H. S. P., Rabipour, Rafi, Chu, Chung-Cheung
• Primary Examiner(s): Voeltz, Emanuel Todd
• Assistant Examiner(s): SOFOCLEOUS, MICHAEL D
• Application Number:
  US08/888,276
• Time in Patent Office: 957 Days
• Field of Search: 704/207, 704/214, 704/219, 704/220, 704/222, 704/223
• US Class Current: 704/201
• CPC Class Codes: G10L 19/012 Comfort noise or silence coding
• CPC Class Codes: G10L 19/06 Determination or coding of the spectral characteristics, e.g. of the short-term prediction coefficients

Abstract

The present invention relates to methods and devices for processing data frames representative of audio information in digitized and compressed form. The method comprises the steps of classifying successive data frames into frames containing speech sounds and non-speech sounds, altering parameters of the data frames identified as containing non-speech sounds for eliminating or at least substantially reducing artifacts that distort the acoustic background noise. In addition, the data frame identified as containing non-speech sounds are low-pass filtered. Finally, a signal level compensation is effected to avoid undesired fluctuations in the signal level.

Claims

1. A signal processing apparatus, comprising:

a) an input for receiving a signal derived from audible sound, the signal conveying a plurality of successive data frames, each data frame being representative of audio information in digitized and compressed form, each data frame including;

a coefficient segment;

an excitation segment;

b) an output;

c) a detector coupled to said input for distinguishing data frames containing speech sounds from data frames containing non-speech sounds;

d) a noise conditioning device;

e) a selector device capable of acquiring two operative conditions, namely a first operative condition and a second operative condition, said selector device being responsive to said detector for switching between the two operative conditions, when said detector distinguishes a data frame as containing speech sounds said selector acquiring the first operative condition, in said first operative condition said selector device causing transfer of a data frame to said output substantially without altering the coefficient segment of the data frame, when said detector distinguishes a data frame as containing non-speech sounds said selector acquiring the second operative condition, to transfer the data frame to said noise conditioning device,

said noise conditioning device being operative for processing the coefficient segment of the data frame received by the noise conditioning device in dependence upon parameters of preceding data frames applied to said input to derive a noise conditioned coefficient segment, the noise conditioned coefficient segment having an impulse response being characterized by a first frequency domain behavior, said noise conditioning device being further operative for low pass filtering the impulse response of the noise conditioned coefficient segment to derive an output coefficient segment having an impulse response characterized by a second frequency domain behavior different from said first frequency domain behavior, said noise conditioning device being further operative to transfer the output coefficient segment to said output.

2. A signal processing apparatus as defined in claim 1, wherein said noise conditioning device further comprises:

a noise conditioning unit for processing a coefficient segment of the data frame received by the noise conditioning device to derive a noise conditioned coefficient segment;

an impulse response computing unit for processing said noise conditioned coefficient segment to derive the impulse response characterized by the first frequency domain behavior;

a low-pass filter for low pass filtering the impulse response characterized by a first frequency domain behavior to derive the impulse response characterized by the second frequency domain behavior;

an auto-correlation unit for processing the impulse response characterized by the second frequency domain behavior to derive the output coefficient segment.

3. A signal processing apparatus as defined in claim 2, wherein said low pass filter is operative to process the impulse response characterized by the first frequency domain behavior for attenuating frequencies above a certain threshold in the impulse response characterized by the first frequency domain behavior to derive the impulse response characterized by the second frequency domain behavior.

4. A signal processing apparatus as defined in claim 3, wherein said certain threshold is about 3500 Hz.

5. A signal processing apparatus as defined in claim 1, wherein the data frame includes a data element indicative of a signal energy, said noise conditioning device comprises a signal level correction unit for selectively altering the data element indicative of a signal energy.

6. A signal processing apparatus as defined in claim 5, wherein said signal level correction unit is operative for comparing the coefficient segment received by the noise conditioning device and the output coefficient segment to derive a correction factor, the correction factor being indicative of a degree of variation between the coefficient segment received by the noise conditioning device and the output coefficient segment.

7. A signal processing apparatus as defined in claim 6, wherein said signal level correction unit alters the data element indicative of a signal energy on a basis of the correction factor.

8. A signal processing apparatus as defined in claim 1, wherein said noise conditioning device is operative for calculating a noise conditioned coefficient segment on a basis of the coefficient segments of preceding data frames applied to said input.

9. A signal processing apparatus as defined in claim 8, wherein number of said preceding data frames is about 19.

10. A signal processing apparatus as defined in claim 1, wherein said noise conditioning device processes the data frame containing non-speech sounds substantially without synthesizing an audio signal conveyed by the data frame.

11. A signal processing apparatus as defined in claim 1, wherein said apparatus is suitable for use in a radio frequency communication system comprising:

a first mobile terminal;

a second mobile terminal;

a base station functionally associated to said first mobile terminal and said second mobile terminal.

12. A method for serially reducing background noise artifacts in a signal derived from audible sound, the signal conveying a succession of data frames, each data frame being representative of audio information in digitized and compressed form, each data frame including a coefficient segment and an excitation segment, said method comprising:

a) receiving the signal derived from audible sound;

b) classifying each data frame in the signal as containing either one of speech sounds and non-speech sounds;

c) transferring the data frames classified as containing speech sounds to an output;

d) processing each frame classified as containing non-speech sounds to alter the coefficient segment thereof in dependence of coefficient segments of preceding data frames to effect a reduction in background noise artifacts in the frame classified as containing non-speech sounds to derive a noise conditioned coefficient segment, the noise conditioned coefficient segment having an impulse response being characterized by a first frequency domain behavior;

e) low pass filtering the impulse response characterized by the first frequency domain behavior of the noise conditioned coefficient segment to derive an output coefficient segment having an impulse response characterized by a second frequency domain behavior different from said first frequency domain behavior;

f) upon completion of the processing at steps d and e, transferring the data frame with an output coefficient segment to said output.

13. A method as defined in claim 12, wherein the data frame includes a data element indicative of a signal energy, said method further comprising selectively altering the data element indicative of a signal energy.

14. A method as defined in claim 13, further comprising comparing the coefficient segment of the frame classified as containing non-speech sounds and the output coefficient segment to derive a correction factor, the correction factor being indicative of a degree of variation between the coefficient segment of the frame classified as containing non-speech sounds and the output coefficient segment.

15. A method as defined in claim 14, the data element indicative of a signal energy is altered on a basis of the correction factor.

16. A method as defined in claim 12, comprising calculating a new coefficient segment for a data frame classified as containing non-speech sounds on a basis of coefficient segments of preceding data frames.

17. A method as defined in claim 16, comprising:

calculating an average of the coefficient segments in the current data frame classified as containing non-speech sounds and the preceding data frames;

replacing the coefficient segment of the current data frame classified as containing non-speech sounds with the average of coefficient segments.

18. A method as defined in claim 12, further comprising:

processing the noise conditioned coefficient segment to derive the impulse response characterized by the first frequency domain behavior;

processing the impulse response characterized by the second frequency domain behavior on the basis of an auto-correlation computation to derive the output coefficient segment.

19. A method as defined in claim 12, wherein low pass filtering the impulse response characterized by the first frequency domain behavior of the noise conditioned coefficient segment attenuates frequencies above a certain threshold in an audio signal synthesized on the basis of the data frame.

20. A communication system including:

a) an encoder including an input for receiving a signal derived from audible sound, said encoder being operative to convert the signal into a succession of data frames representative of audio information in digitized and compressed form, each data frame including a coefficient segment and an excitation segment;

b) a decoder remote from said encoder, said decoder including an input for receiving data frames representative of audio information in digitized and compressed form to convert the data frames into an audio signal;

c) a communication path between said encoder and said decoder, said communication path allowing data frames generated by said encoder to be transported to the input of said decoder;

d) a signal processing apparatus in said communication path for reducing background noise artifacts in data frames transported from said encoder toward said decoder, said signal processing apparatus comprising;

an input for receiving the succession of data frames from said encoder;

an output for issuing a succession of data frames toward the input of said decoder;

a detector coupled to the input of said signal processing apparatus for distinguishing data frames containing speech sounds from data frames containing non-speech sounds;

a noise conditioning device;

a selector device capable of acquiring two operative conditions, namely a first operative condition and a second operative condition, said selector device being responsive to said detector for switching between the two operative conditions, when said detector distinguishes a data frame as containing speech sounds said selector acquiring the first operative condition, in said first operative condition said selector device causing transfer of a data frame to said output substantially without altering the coefficient segment of the data frame, when said detector distinguishes a data frame as containing non-speech sounds said selector acquiring the second operative condition, to transfer the data frame to said noise conditioning device,

said noise conditioning device being operative for processing the coefficient segment of the data frame received by the noise conditioning device in dependence upon parameters of preceding data frames applied to said input to derive a noise conditioned coefficient segment, the noise conditioned coefficient segment having an impulse response being characterized by a first frequency domain behavior, said noise conditioning device being further operative for low pass filtering the impulse response of the noise conditioned coefficient segment to derive an output coefficient segment having an impulse response characterized by a second frequency domain behavior different from said first frequency domain behavior, said noise conditioning device being further operative to transfer the output coefficient segment to said output.