Feedback control system for telecommunications systems

• US 5,978,783 A
• Filed: 06/25/1997
• Issued: 11/02/1999
• Est. Priority Date: 01/10/1995
• Status: Expired due to Term

• Current Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Weaver, Carl F., Tarraf, Ahmed A., Meyers, Martin H.
• Primary Examiner(s): Hafiz, Tariq R.
• Assistant Examiner(s): Starks, Wilbert L.
• Application Number:
  US08/882,436
• Time in Patent Office: 860 Days
• Field of Search: 706/15, 706/14, 704/203
• US Class Current: 706/21
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 19/26 Pre-filtering or post-filtering
• CPC Class Codes: G10L 25/30 using neural networks
• CPC Class Codes: G10L 25/69 for evaluating synthetic or decoded voice signals
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 3/22 Arrangements for supervision, monitoring or testing

Abstract

Telecommunications processing is applied to a reference signal to generate a signal under test. A fidelity measure is generated characterizing the fidelity of the signal under test relative to the reference signal. A control signal is generated from the fidelity measure, where the control signal is used as a feedback signal to adjust the telecommunications processing. In one embodiment, the reference signal is a speech signal and the signal under test is a decoded speech signal generated by encoding, transmitting, and decoding the reference speech signal. The fidelity signal is an average mean opinion score (MOS) and the control signal is used to control the speech decoding processing. For example, the speech decoding processing may involve a speech decoder followed by a post filter, and the control signal is the cut-off frequency of the post filter. Such an embodiment may be applied to real-time speech processing to adjust the speech decoding processing in the receiver of a telecommunications system that includes a speech encoder in a transmitter and a communications network. Intermittently, the transmitter encodes reference signals that are also known to the receiver. The receiver performs the processing of the present invention to update its speech decoding processing to optimize speech quality.

Claims

1. A method for processing telecommunications signals, comprising the steps of:

(a) applying telecommunications processing to generate a signal under test corresponding to a reference signal, wherein the telecommunications processing comprises transmission over a communications (comm) network;

(b) generating a fidelity measure characterizing fidelity of the signal under test relative to the reference signal, wherein the telecommunications processing has not been applied to the reference signal;

(c) generating a control signal from the fidelity measure; and

(d) using the control signal as a feedback signal to adjust the telecommunications processing.

2. The invention of claim 1, wherein the reference signal is an input speech signal and the signal under test is a decoded speech signal.

3. The invention of claim 2, wherein:

the telecommunications processing comprises a speech encoder, the comm network, and a speech decoder; and

step (a) comprises the steps of;

(1) generating an encoded speech signal by applying the speech encoder to the input speech signal; and

(2) generating the decoded speech signal by applying the comm network and the speech decoder to the encoded speech signal.

4. The invention of claim 2, wherein the control signal is used as the feedback signal to adjust speech decoding processing.

5. The invention of claim 2, wherein:

step (a) comprises the step of applying speech decoding processing to generate the decoded speech signal from a received signal corresponding to a reference speech signal that has been encoded by a speech encoder and transmitted over the comm network;

step (b) comprises the step of generating an average mean opinion score (MOS) from the decoded speech signal and the reference speech signal;

step (c) comprises the step of generating the control signal from the average MOS; and

step (d) comprises the step of using the control signal to adjust the speech decoding processing.

6. The invention of claim 5, wherein:

the speech decoding processing is implemented by a speech decoder followed by a post filter; and

the control signal is a cut-off frequency for the post filter.

7. The invention of claim 6, wherein:

a performance measure is generated from a difference between the average MOS and a desired MOS; and

an adjusted cut-off frequency is generated from the performance measure using a gradient descent rule.

8. The invention of claim 5, wherein steps (a)-(d) are applied intermittently to one or more sets of reference signal and signal under test during real-time speech processing.

9. The invention of claim 1, wherein step (b) comprises the steps of:

(1) generating a feature set from the reference signal and the signal under test; and

(2) applying an AI tool to the feature set to generate the fidelity measure.

10. The invention of claim 1, wherein:

the telecommunications processing comprises a signal encoder, the comm network, and a signal decoder; and

step (a) comprises the steps of;

(1) generating an encoded signal by applying the signal encoder to the reference signal; and

(2) generating the signal under test by applying the comm network and the signal decoder to the encoded signal.

11. The invention of claim 1, wherein steps (a)-(d) are applied two or more times to one or more sets of reference signal and signal under test.

12. The invention of claim 1, wherein:

the telecommunications processing comprises a signal decoder;

step (a) comprises the step of generating the signal under test by applying the signal decoder to a received signal corresponding to the reference signal that has been encoded by a signal encoder and transmitted over the comm network; and

the control signal is used to adjust processing of the signal decoder.

13. The invention of claim 12, wherein steps (a)-(d) are applied intermittently to one or more sets of reference signal and signal under test during real-time signal processing.

14. An apparatus for processing telecommunications signals, comprising:

(a) means for applying telecommunications processing to generate a signal under test corresponding to a reference signal, wherein the telecommunications processing comprises transmission over a comm network;

(b) means for generating a fidelity measure characterizing fidelity of the signal under test relative to the reference signal, wherein the telecommunications processing has not been applied to the reference signal;

(c) means for generating a control signal from the fidelity measure; and

(d) means for using the control signal as a feedback signal to adjust the telecommunications processing.

15. A system for processing telecommunications signals, comprising:

(a) a telecommunications system adapted to generate a signal under test from a reference signal, wherein the telecommunications system comprises transmission over a comm network;

(b) a fidelity measuring apparatus adapted to generate a fidelity measure characterizing fidelity of the signal under test relative to the reference signal, wherein the telecommunications system has not processed the reference signal; and

(c) a control signal generator adapted to generate a control signal from the fidelity measure, wherein the control signal is useable as a feedback signal to adjust the telecommunications system.

16. The invention of claim 15, wherein the reference signal is an input speech signal and the signal under test is a decoded speech signal.

17. The invention of claim 16, wherein:

the telecommunications system comprises a speech encoder, the comm network, and a speech decoder;

the speech encoder are adapted to generate an encoded speech signal from the input speech signal; and

the comm network and the speech decoder are adapted to generate the decoded speech signal from the encoded speech signal.

18. The invention of claim 16, wherein the control signal is used as the feedback signal to adjust speech decoding processing.

19. The invention of claim 16, wherein:

the telecommunications system is adapted to apply speech decoding processing to generate the decoded speech signal from a received signal corresponding to a reference speech signal that has been encoded by a speech encoder and transmitted over the comm network;

the fidelity measuring apparatus is adapted to generate an average MOS from the decoded speech signal and the reference speech signal;

the control signal generator is adapted to generate the control signal from the average MOS; and

the telecommunications system uses the control signal to adjust the speech decoding processing.

20. The invention of claim 19, wherein:

the telecommunications system comprises a speech decoder followed by a post filter; and

the control signal is a cut-off frequency for the post filter.

21. The invention of claim 20, wherein:

the control signal generator generates a performance measure from a difference between the average MOS and a desired MOS; and

the control signal generator generates an adjusted cut-off frequency from the performance measure using a gradient descent rule.

22. The invention of claim 19, wherein processing of the telecommunications system, the fidelity measuring apparatus, and the control signal generator is applied intermittently to one or more sets of reference signal and signal under test during real-time speech processing.

23. The invention of claim 15, wherein the fidelity measuring apparatus:

(1) generates a feature set from the reference signal and the signal under test; and

(2) applies an AI tool to the feature set to generate the fidelity measure.

24. The invention of claim 15, wherein:

the telecommunications system comprises a signal encoder, the comm network, and a signal decoder;

the signal encoder generates an encoded signal by applying the signal encoder to the reference signal; and

the comm network and the signal decoder generate the signal under test from the encoded signal.

25. The invention of claim 15, wherein processing of the telecommunications system, the fidelity measuring apparatus, and the control signal generator is applied two or more times to one or more sets of reference signal and signal under test.

26. The invention of claim 15, wherein:

the telecommunications system comprises a signal decoder;

the signal decoder generates the signal under test from a received signal corresponding to the reference signal that has been encoded by a signal encoder and transmitted over the comm network; and

the control signal is used to adjust processing of the signal decoder.

27. The invention of claim 26, wherein processing of the telecommunications system, the fidelity measuring apparatus, and the control signal generator is applied intermittently to one or more sets of reference signal and signal under test during real-time signal processing.