Audio signal quality enhancement in a digital network

• US 20040076271A1
• Filed: 10/27/2003
• Published: 04/22/2004
• Est. Priority Date: 12/29/2000
• Status: Active Grant

• Current Assignee: Nokia Solutions and Networks US LLC (Nokia Corporation)
• Original Assignee: Nokia Solutions and Networks US LLC (Nokia Corporation)
• Inventors: Koistinen, Tommi, Kirla, Olli
• Granted Patent: US 7,539,615 B2
• Time in Patent Office: Days
• US Class Current: 379/88.11
• CPC Class Codes: G10L 19/005 Correction of errors induced by the transmission channel, if related to the coding algorithm
• CPC Class Codes: G10L 2021/02082 the noise being echo, reverberation of the speech
• CPC Class Codes: G10L 21/0208 Noise filtering
• CPC Class Codes: G10L 25/69 for evaluating synthetic or decoded voice signals
• CPC Class Codes: H04B 3/23 using a replica of transmitted signal in the time domain, e.g. echo cancellers

Abstract

The invention relates to a network element (1) and a method for enhancing the quality of digitised analogue signals transmitted in parameterised coded form via a digital network. In order to enable an enhancement of the quality of the digitised analogue signals on network side, the network element comprises means (20, 21) for extracting signals from and insert signals into the network, first processing means (24) for processing the extracted parameters in the parameter domain with functions suitable to enhance the quality of the digitised analogue signals and second processing means (26) for processing the extracted parameters in the linear domain with functions suitable to enhance the quality of the digitised analogue signals. Moreover included analysing and selecting means (23, 27) determine the expected enhancement of quality in the different processing domains and cause a corresponding insertion of processed signals back into the network. The proposed method comprises corresponding steps.

Claims

1. Network element (1) for enhancing the quality of digitised analogue signals transmitted at least in parameterised coded form via a digital network to which the network element (1) has access, comprising

a payload extraction block (20,30,40) for extracting coded digitised analogue signals from the digital network, which coded digitised analogue signals include at least in part parameterised coded digitised analogue signals;

first processing means (24,34,44) for processing the extracted parameterised coded digitised analogue signals in the parameter domain with functions suitable to enhance the quality of the digitised analogue signals;

second processing means (26,36,46) for processing at least part of the extracted coded digitised analogue signals in the linear domain with functions suitable to enhance the quality of the digitised analogue signals;

a payload insertion block (21,31,41) for inserting processed coded digitised analogue signals to the digital network; and

analysing and selecting means (23,27,33,37) for determining the quality improvement of the digitised analogue signals resulting from a processing of the extracted coded digitised analogue signals in the parameter domain and from a processing of the extracted coded digitised analogue signals in the linear domain and for causing that at least coded digitised analogue signals processed by the processing means (24,26,34,36) leading to the better improvement are inserted back to the digital network by the payload insertion block (21,31,41).

2. Network element (1) according to claim 1, wherein the functions for processing the parameterised coded digitised analogue signals by the first processing means (24,34,44) include echo cancellation, noise reduction, and/or level control.

3. Network element (1) according to one of the preceding claims, wherein the functions for processing coded digitised analogue signals by the second processing means (26,36,46) include echo cancellation, noise reduction, level control, transcoding and/or speech mixing.

4. Network element (1) according to one of the preceding claims, wherein the analysing and selecting means are suited for analysing the digitised analogue signals before and after processing by the first and the second processing means for determining the better quality improvement.

5. Network element (1) according to one of the preceding claims, wherein the analysing and selecting means comprises a neural network for determining the better quality improvement.

6. Network element (1) according to one of the preceding claims, characterised by a bad frame handler (22,32,42) for detecting in the extracted coded digitised analogue signals missing frames and/or frames in disorder, and for regenerating missing frames and/or reordering frames in disorder in the extracted signals.

7. Network element (1) according to one of the preceding claims, characterised by analysing means (23,33) for determining whether any processing is to be applied to the extracted parameterised coded digitised analogue signals and for selecting the functions that are to be applied to the extracted coded digitised analogue signals by the first processing means (24,34) and/or the second processing means (26,36) depending on the extracted signals and/or an external control signal.

8. Network element (1) according to one of the preceding claims, characterised by a control block (29,39,49) for receiving an external control signal and for controlling the selection of the processing applied to the extracted coded signals directly or via analysing means (23,33,37).

9. Network element (1) according to one of the preceding claims, wherein the payload extraction block (20) is suitable to extract parameterised coded digitised analogue signals from the IP stack of a packet-based network and wherein the payload insertion block (21) is suitable to insert parameterised coded digitised analogue signals to the IP stack of said packet-based network.

10. Network element (1) according to claim 9, wherein the network element (1) further comprises:

decoding means (25) for decoding said parameterised coded digitised analogue signals and for forwarding the decoded signals to the second processing means (26);

encoding means (28) for encoding the signals processed and output by the second processing means (26);

first selecting means (23) for receiving the extracted coded digitised analogue signals from the payload extraction block (20) and for forwarding the extracted coded digitised analogue signals either to the first processing means (24) or to the second processing means (26) via the decoding means (25);

second selecting means (27) for receiving as inputs the outputs of the first processing means (24) and of the second processing means (26) and for selecting which output is to be forwarded to the insertion payload block (21); and

analysing means (23) for determining whether the extracted coded digitised analogue signals are to be processed by the first processing means (24) or the second processing means (26) and for controlling the first selecting means (23) and the second selecting means (27) accordingly.

11. Network element according to one of claim 1 to 8, wherein the payload extraction block (30,40) is suitable for extracting a TFO stream and a PCM stream from a TDM based network, for separating said TFO stream from said PCM stream and for providing the TFO stream as parameterised coded digitised analogue signals, and wherein the payload insertion block (31,41) is suitable for combining a TFO stream with a PCM stream and for inserting the combined stream, or only a PCM stream if no TFO stream is provided to the payload insertion block (31,41), to the TDM based network.

12. Network element according to claim 11, wherein the network element further comprises:

decoding means (35) for decoding said TFO stream and for forwarding the decoded signals to the second processing means (36);

encoding means (38) for encoding the signals processed and output by the second processing means (36);

means (33) for receiving the TFO stream from the payload extraction block (30) and for forwarding the TFO stream to the first processing means (34) and via the decoding means (35) to the second processing means (36);

selecting means (37) for receiving as inputs the outputs of the first processing means (34) and of the encoding means (38) and for forwarding the output with the better quality improvement of the digitised analogue signal to the insertion payload block (31);

analysing means (33) for determining whether a processing in the first processing means (34) or in the second processing means (36) results in a better quality improvement of the digitised analogue signals and for controlling the selecting means (37) accordingly; and

means (31) for forming a PCM stream out of the output of the second processing means (36) and for forwarding said PCM stream to the payload insertion block (31).

13. Network element according to claim 12, wherein the payload extraction block (30,40) is suitable for providing in addition the PCM stream as non-parameterised coded digitised analogue signals, further comprising means (33) for forwarding the extracted PCM stream to the second processing means (36) and for forwarding the output of the second processing means (36) to the payload insertion block (31), if no TFO stream is available for processing.

14. Network element according to one of claims 11, wherein the payload extraction block (30,40) is suitable for providing in addition the PCM stream as non-parameterised coded digitised analogue signals, wherein the TFO stream is processed in the first processing means (44) in the parameter domain and the PCM stream is processed in the second processing means (46) in the linear domain, and wherein at least the processed PCM stream is forwarded to the payload insertion block (41).

15. Network element (1) according to one of the preceding claims, wherein the first processing means (24,34,44) is suited for comparing gain parameters of the extracted parameterised coded digitised analogue signals with a desired gain, for forming corresponding new gain parameters and for replacing the original gain parameters by the new gain parameters in the extracted parameterised coded digitised analogue signals.

16. Network element (1) according to claim 15, wherein the first processing means (24,34,44) comprises:

decoding means (50) for linearising extracted parameterised coded digitised analogue signals and for providing decoded gain parameters of the digitised analogue signals;

means (51) for estimating the level of the linearised coded digitised analogue signal and for determining desired gain values based on the estimated level of the digitised analogue signal and a desired target level of the digitised analogue signal;

means (52) for determining out of decoded gain parameters of the coded digitised analogue signal and the desired gain values new gain parameters suitable for achieving the desired gain by linear-to-parameter domain mapping; and

means (53) for re-quantising the new gain parameters and for replacing the original gain parameters with the new gain parameters in the parameterised coded digitised analogue signal.

17. Network element (1) according to claim 16, wherein the means (51) for estimating the level of the digitised analogue signal comprises a voice activity detector for ensuring that only speech signals are estimated in the signal level estimate.

18. Network element (1) according to one of the preceding claims, wherein the first processing means (24,34,44) is suited for attenuating noise portions and low level signal portions of the extracted parameterised coded digitised analogue signals in the time domain and for correspondingly replacing the gain parameters in the extracted parameterised coded digitised analogue signals.

19. Network element (1) according to one of the preceding claims, wherein the first processing means (24,34,44) is suited for attenuating frequency portions of noise in the extracted parameterised coded digitised analogue signals which have approximately the same energy as a noise estimate and for correspondingly replacing linear prediction parameters in the extracted parameterised coded digitised analogue signals.

20. Network element (1) according to claim 19, wherein the first processing means (24,34,44) comprises:

a decoder (60) for decoding linear prediction coefficients from extracted coded digitised analogue signals;

means (64) for estimating a long term power spectrum of the noise of the digitised analogue signals;

means (65) for estimating a short-term power spectrum of the noise of the digitised analogue signals;

means (67) for determining a desired spectrum depending on the difference between the long term spectrum and the short term spectrum;

means (68) for determining new linear prediction coefficients according to the desired spectrum;

means (69) for quantising the new linear prediction coefficients parameters or for converting them to line spectral pairs parameters and for replacing them for the old parameters in the extracted parameterised coded digitised analogue signal.

21. Network element (1) according to one of the preceding claims, wherein the payload extraction block (20,30,40) is suited for extracting in addition further coded digitised analogue signals from the digital network transmitted in the opposite direction as compared to the extracted coded digitised analogue signals, which further coded digitised analogue signals include at least in part parameterised coded digitised analogue signals;

and wherein the first processing means (24,34,44) is suited for comparing the first and the further parameterised coded digitised analogue signals in order to detect echoes in the first parameterised coded digitised analogue signals and for replacing portions of the first parameterised coded digitised analogue signal by comfort noise portions, if an echo was determined in the portion of the first parameterised coded digitised analogue signal.

22. Network element (1) according to claim 21, wherein the first processing means (24,34,44) comprises:

a first decoder (70) for linearising extracted coded digitised analogue signals from a first direction;

a second decoder (71) for linearising the further extracted coded digitised analogue signals from an opposite direction;

an echo analysis block (72) for detecting an echo in a portion of the first extracted signal and

means (72-76) for generating comfort noise and for replacing the original portion of the first extracted signal with a corresponding comfort noise parameter in case an echo was detected.

23. Network element (1) according to one of the preceding claims, wherein the payload extraction block (20,30,40) is suited for extracting in addition further coded digitised analogue signals from the digital network transmitted in the opposite direction as compared to the extracted coded digitised analogue signals, which further coded digitised analogue signals include at least in part parameterised coded digitised analogue signals;

and wherein the first processing means (24,34,44) is suited for attenuating an echo signal in the first parameterised coded digitised analogue signals making use of the further parameterised coded digitised analogue signals and for suppressing the residual echo signal.

24. Network element (1) according to claim 23, wherein the first processing means (24,34,44) comprises:

a first decoder (80) for linearising extracted coded digitised analogue signals from a first direction;

a second decoder (81) for linearising the further extracted coded digitised analogue signals from an opposite direction;

an adaptive filter (82) and a control logic (88) receiving linearised signals from the first and the second decoder (80,81) for attenuating echo signals in the linearised signal received from the first decoder (80);

a non linear processor (84) for residual echo suppression based on residual echo signals received from the adaptive filter (82) and on a noise estimation of the linearised signals from the first direction; and

means (83-86,88) for generating comfort noise based on the residual echo suppression and for replacing the original portion of the first extracted signal with a corresponding comfort noise parameter in case an echo was detected.

25. Network element (1) according to one of claims 21 to 24, wherein the first processing means (24,34,44) is suited for by-passing the first parameterised coded digitised analogue signals without processing, if there is no signal activity in the opposite direction or if the signal level of the extracted parameterised coded digitised analogue signals is below a threshold level in the opposite direction.

26. Network element (1) according to one of the preceding claims, wherein the extracted coded digitised analogue signals are coded speech or coded video.

27. Method for enhancing the quality of digitised analogue signals transmitted at least in parameterised coded form via a digital network, comprising:

extracting coded digitised analogue signals from the digital network, which coded digitised analogue signals include at least in part parameterised coded digitised analogue signals;

determining the quality improvement of the digitised analogue signals to be expected by a processing of the extracted encoded digitised analogue signals in the parameter domain and by a processing of the extracted encoded digitised analogue signals in the linear domain;

processing the extracted parameterised coded digitised analogue signals in the parameter domain at least if a greater quality improvement is expected by processing in the parameter domain, with functions suitable for enhancing the quality of digitised analogue signals; and

processing at least part of the extracted coded digitised analogue signals in the linear domain at least if a greater quality improvement is expected by processing in the linear domain, with functions suitable for enhancing the quality of digitised analogue signals; and

inserting at least those processed coded digitised analogue signals to the digital network that were processed in the domain, the processing in which was expected to result in a greater quality improvement.

28. Method according to claim 27, further comprising:

decoding the extracted parameterised coded digitised analogue signals for processing in the linear domain; and

encoding the processed decoded signals after processing in the linear domain to form parameterised coded digitised analogue signals again.

29. Method according to claim 28, further comprising:

transforming the processed decoded signals to form non-parameterised coded digitised analogue signals; and

inserting the non-parameterised coded digitised analogue signals to the digital network.

30. Method according to claim 27, wherein part of the extracted encoded digitised analogue signals are non-parameterised coded digitised analogue signals corresponding to the extracted parameterised coded digitised analogue signals, wherein the parameterised coded digitised analogue signals are processed in the parameter domain if a greater quality improvement is expected by processing in the parameter domain, wherein the non-parameterised coded digitised analogue signals are processed in the linear domain, wherein the processed non-parameterised coded digitised analogue signals are inserted to the digital network again and wherein the processed parameterised coded digitised analogue signals are inserted to the digital network again if a greater quality improvement is expected by processing in the parameter domain.

31. Method according to one of claims 27 to 30, wherein the quality improvement of a processing in the linear and in the parameter domain is determined by analysing the digitised analogue signal before and after processing in the linear and in the parameter domain.

32. Method according to one of claims 27 to 31, wherein the quality improvement of a processing in the linear and in the parameter domain is determined by using a neural network.

33. Method according to one of claims 27 to 32, further comprising:

selecting those processing functions that are suitable for an enhancement of the quality of the digitised analogue signals in question and carrying out only those processing functions.

34. Method according to one of claims 27 to 33, wherein the processing in the parameter domain comprises for a gain control comparing gain parameters of the extracted parameterised coded digitised analogue signals with a desired gain for forming corresponding new gain parameters and for replacing the original gain parameters by the new gain parameters in the extracted parameterised coded digitised analogue signals.

35. Method according to claim 34, comprising

linearising extracted parameterised coded digitised analogue signals and providing decoded gain parameters of the digitised analogue signals;

estimating the level of the linearised coded digitised analogue signals and determining desired gain values based on the estimated signal level and a desired target signal level;

determining out of decoded gain parameters of the coded digitised analogue signals and the desired gain values new gain parameters suitable for achieving the desired gain by linear-to-parameter domain mapping; and

re-quantising the new gain parameters and replacing the original gain parameters with the new gain parameters in the coded digitised analogue signals signal.

36. Method according to one of claims 27 to 35, wherein the processing in the parameter domain comprises for noise suppression attenuating noise portions and low level signal portions of the extracted parameterised coded digitised analogue signals in the time domain and for correspondingly replacing gain parameters in the extracted parameterised coded digitised analogue signals.

37. Method according to one of claims 27 to 36, wherein the processing in the parameter domain comprises for noise suppression attenuating frequency portions of noise in the extracted parameterised coded digitised analogue signals which have which have approximately the same energy as a noise estimate and for correspondingly replacing linear prediction parameters in the extracted parameterised coded digitised analogue signals.

38. Method according to claim 37, comprising

decoding linear prediction coefficients from extracted coded digitised analogue signals;

estimating a long term power spectrum of the noise of the digitised analogue signals;

estimating a short-term power spectrum of the noise of the digitised analogue signals;

determining a desired spectrum in dependence of the difference between the long term spectrum and the short term spectrum;

determining new linear prediction coefficients according to the desired spectrum; and

quantising the new linear prediction coefficients parameters or converting them to line spectral pairs parameters and replacing them for the old parameters in the parameterised coded digitised analogue signal.

39. Method according to one of claims 27 to 38, wherein the processing in the parameter domain comprises for echo suppression extracting in addition further parameterised coded digitised analogue signals transmitted in the opposite direction, comparing the first and the further parameterised coded digitised analogue signals in order to detect echoes in the first parameterised coded digitised analogue signals and replacing portions of the first parameterised coded digitised analogue signal by generated portions with comfort noise parameters, if an echo was determined in the portion of the first parameterised coded digitised analogue signal.

40. Method according to claim 39, comprising linearising the extracted coded digitised analogue signals from the first and the opposite direction before comparing them.

41. Method according to one of claims 27 to 40, comprising extracting in addition further coded digitised analogue signals from the digital network transmitted in the opposite direction as compared to the extracted coded digitised analogue signals, which further coded digitised analogue signals include at least in part parameterised coded digitised analogue signals, attenuating an echo signal in the first parameterised coded digitised analogue signals making use of the further parameterised coded digitised analogue signals and suppressing the residual echo signal.

42. Method according to claim 41, comprising linearising the extracted coded digitised analogue signals from both directions before attenuating the echo signal, generating comfort noise based on the result of the suppression and an estimated noise in the first digitised analogue signal and replacing a portion of the original first extracted signal in which an echo was detected with a portion comprising a corresponding comfort noise parameter.

43. Method according to one of claims 39 to 42, wherein the processing in the parameter domain comprises by-passing the first parameterised coded digitised analogue signals without echo detection, if there is no signal activity in the opposite direction or if the signal level of the extracted parameterised coded digitised analogue signals is below a threshold level in the opposite direction.