Signature noise removal

• US 8,271,279 B2
• Filed: 11/30/2006
• Issued: 09/18/2012
• Est. Priority Date: 02/21/2003
• Status: Active Grant

• Current Assignee: Malikie Innovations Limited (Key Patent Innovations Limited)
• Original Assignee: QNX Software Systems Limited (Canada) (Blackberry Limited)
• Inventors: Hetherington, Phillip A., Paranjpe, Shreyas A.
• Primary Examiner(s): He, Jialong
• Application Number:
  US11/607,340
• Publication Number:
  US 20070078649A1
• Time in Patent Office: 2,119 Days
• Field of Search: 704/226, 704/233, 704/205, 704/211
• US Class Current: 704/233
• CPC Class Codes: G10L 2021/02085 Periodic noise
• CPC Class Codes: G10L 21/0216 characterised by the method used for estimating noise
• CPC Class Codes: G10L 21/0232 Processing in the frequency domain
• CPC Class Codes: G10L 25/45 characterised by the type of analysis window

Abstract

A speech enhancement system improves the perceptual quality of a processed voice signal. The system improves the perceptual quality of a voice signal by removing unwanted noise components from a voice signal. The system removes undesirable signals that may result in the loss of information. The system receives and analyzes signals to determine whether an undesired random or persistent signal corresponds to one or more modeled noises. When one or more noise components are detected, the noise components are substantially removed or dampened from the signal to provide a less noisy voice signal.

Claims

1. A speech enhancement system operative to suppress noise from a received signal comprising:

a background noise estimator that measures a background noise level in the received signal;

a transient noise detector operative to store a model of a noise component within a memory and operative to detect the presence of a transient noise in the received signal; and

a noise attenuator in communication with the transient noise detector and operative to substantially remove the transient noise from the received signal when an attribute of the received signal substantially matches an attribute of the stored model of the noise component;

where the transient noise detector is operative to disable or modulate the background noise estimator during a period of time when an instantaneous background noise level of the received signal exceeds an average background noise level of the received signal by more than a predetermined threshold.

2. The system of claim 1 where the noise detector is operative to compare the attribute of the received signal to the attribute of the stored model of the noise component, and where the transient noise detector comprises circuitry or a computer-readable storage medium that stores instructions executable by a processor to detect the presence of the transient noise in the received signal.

3. The system of claim 2 where the model of the noise component comprises a spectral attribute of the noise component and a temporal attribute of the noise component.

4. The system of claim 3 where the temporal component comprises a first sound event and a substantially similar second sound event separated by a period of time.

5. The system of claim 3 where the spectral component comprises one or more attributes of a spectral shape of a sound event associated with a road noise.

6. The system of claim 3 where the noise detector and the noise attenuator are coupled to a vehicle.

7. The system of claim 1 where the model of the noise component comprises a dynamic model, and where the noise detector changes the dynamic model in response to detection of changing conditions in the received signal.

8. The system of claim 1, where the transient noise detector comprises modeling logic that fits a function to a selected portion of the received signal in a time-frequency domain to evaluate the spectro-temporal shape characteristics of a sound event in the received signal, and where the modeling logic identifies the sound event as a noise event based on a correlation between the function and a signal envelope of the sound event.

9. The system of claim 1, where the noise attenuator is operative to add the stored model of the noise component to a recorded or modeled continuous noise for removal from the received signal when the transient noise is detected in the received signal.

10. A noise detector operative to detect a noise that may affect a signal comprising:

an analog to digital converter operative to convert a received signal into a digital signal;

a windowing function operative to separate the received signal into a plurality of signal analysis windows;

a transform logic operative to transform the plurality of signal analysis windows to the frequency domain; and

a modeling logic operative to store attributes of a noise, and compare the stored attributes to a transformed signal to identify a noise, where the modeling logic fits a function to a selected portion of the transformed signal in a time-frequency domain to evaluate the spectro-temporal shape characteristics of a sound event in the transformed signal, and where the modeling logic identifies the sound event as a noise event based on a correlation between the function and a signal envelope of the sound event.

11. The noise detector of claim 10 where the analog to digital converter converts the received signal into a pulse code modulated signal.

12. The noise detector of claim 10 where the windowing function comprises a Hanning window function generator.

13. The noise detector of claim 10 where the transform module comprises a Fast Fourier Transform logic.

14. The noise detector of claim 13 where the attributes of the noise comprise a temporal characteristic substantially unique to an undesired signal.

15. The noise detector of claim 13 where the attributes of the noise comprise a spectral characteristic substantially unique to an undesired signal.

16. The noise detector of claim 13 where the attributes of the noise comprise temporal characteristics and spectral characteristics substantially unique to an undesired signal.

17. The noise detector of claim 16 where the attributes of the noise comprise spectral shape characteristics of two sound events.

18. The noise detector of claim 17 where the modeling logic is operative to fit a function to a selected portion of the signal in a time-frequency domain to evaluate the spectro-temporal shape characteristics of the two sound events.

19. The noise detector of claim 13 further comprising a residual attenuator operative to track the power spectrum of the received signal.

20. The noise detector of claim 19 where the residual attenuator is operative to limit the transmitted power in a low frequency range to a predetermined threshold when a large increase in signal power is detected.

21. The noise detector of claim 20 where the predetermined threshold is based on the average spectral power of the received signal in the low frequency range from an earlier period in time.

22. The noise detector of claim 10, where the modeling logic comprises circuitry or a computer-readable storage medium that stores instructions executable by a processor to identify the noise.

23. A method operative to substantially remove noises from a signal comprising:

modeling characteristics of a noise to generate a noise model;

analyzing the signal to determine whether characteristics of the signal correspond to characteristics of the noise model;

fitting a function to a selected portion of the signal in a time-frequency domain to evaluate the spectro-temporal shape characteristics of a sound event in the signal;

identifying the sound event as a noise event based on a correlation between the function and a signal envelope of the sound event; and

applying the signal to a noise attenuator that removes characteristics of the sound event from the signal when the sound event is identified as the noise event.

24. The method of claim 23 further comprising modeling a temporal separation between more than one sound event.

25. The method of claim 24 where the spectral shape attributes of the more than one sound event comprises a broadband event with peak energy levels occurring at relatively lower frequencies.

26. The method of claim 23, where the act of applying the signal to the noise attenuator comprises adding the noise model to a recorded or modeled continuous noise for removal from the signal.

27. The method of claim 23, where the noise attenuator comprises circuitry or a computer-readable storage medium that stores instructions executable by a processor to remove the characteristics of the sound event from the signal.