Method and system for implementing a low complexity spectrum estimation technique for comfort noise generation
View Patent ↗A method and system for implementing a low complexity spectrum estimation technique for comfort noise generation are disclosed. Another aspect of the present invention involves segregating filter parameter encoding from an adaptation process for transmission in the form of silence insertion descriptors. A method for implementing a spectrum estimation for comfort noise generation comprises the steps of receiving an input noise signal; approximating a spectrum of the input noise signal using an algorithm over a period of time; detecting an absence of speech signals; and generating comfort noise based on the approximating step when the absence of speech signals is detected; wherein the spectrum of the input noise signal is substantially constant over the period of time.
1. A method for implementing a spectrum estimation for comfort noise generation, the method comprising the steps of:
receiving an input noise component of a signal;
approximating a spectrum of the input noise component using an algorithm over a period of time;
detecting an absence of speech signals
generating comfort noise based on the approximating step when the absence of speech signals is detected; and
performing an internal check to ascertain that the input noise component is within approximately 6 dB of a noise floor;
wherein the spectrum of the input noise component is substantially constant over the period of time.
2. The method of claim 1 , wherein the step of approximating further comprises the step of: shaping the input noise to a spectrum of a predicted signal using an inverse predictor.
3. The method of claim 1 , wherein approximating to at least one of noise spikes and speech segments is prevented.
4. The method of claim 1 , wherein the algorithm is a least mean square algorithm.
5. The method of claim 1 , wherein the algorithm is a leaky least mean square algorithm.
6. The method of claim 1 , wherein the algorithm is a normalized least mean square algorithm.
7. The method of claim 1 , wherein the algorithm is a linear predictive coding algorithm.
8. The method of claim 1 , wherein the generated comfort noise is substantially uniform.
9. The method of claim 1 , further comprising the step of:
normalizing the algorithm for making the approximating step substantially independent of signal amplitude variations.
10. The method of claim 1 , further comprising the step of:
segregating filter parameter encoding into at least one reflection coefficients from the approximating step for transmitting at least one silence insertion descriptor.
11. The method of claim 10 , wherein interoperability between systems is enhanced.
12. The method of claim 11 , wherein MIPS and memory are efficiently utilized.
13. The method of claim 1 , wherein the step of approximating further comprises the step of:
filtering the input noise component by a synthesis filter.
14. The method of claim 13 , wherein the synthesis filter is defined as follows:
H
(
z
)
=
1
∑
i
=
0
M
w
i
z
-
i
wherein M represents a number of taps, w represents a predictor coefficient and H is a function of variable z.
15. The method of claim 13 , wherein the synthesis filter is a 10 th order synthesis filter.
16. The method of claim 1 , wherein the step of approximating further comprises the steps of:
detecting noise between speech data;
adapting to the noise; and
creating silence insertion descriptors based on the adapting step when speech is inactive.
17. The method of claim 16 , wherein the silence insertion descriptors are generated by converting at least one direct form coefficients to at least one reflection coefficients as represented by:
A
m
-
1
(
z
)
=
[
A
m
(
z
)
-
B
m
(
z
)
k
m
]
[
1
-
k
m
2
]
m
=
M
-
1
,
M
-
2
,
…
,
1
B m ( z )= z −m A m ( z −1 )
18. The method of claim 16 , wherein silence insertion descriptors are decoded by converting at least one reflection coefficients to direct form coefficients as represented by:
A
0
(
z
)
=
B
0
(
z
)
=
1
A
m
(
z
)
=
A
m
-
1
(
z
)
+
z
-
1
B
m
-
1
(
z
)
k
m
m
=
1
,
2
,
…
,
M
-
1
B
m
(
z
)
=
z
-
m
A
m
(
z
-
1
)
m
=
1
,
2
,
…
,
M
-
1
19. The method of claim 1 , wherein the step of approximating further comprises the steps of:
detecting noise between speech data;
adapting to the noise; and
generating enhanced noise based on an average spectrum of the input noise component when speech is inactive.
20. A method for implementing a spectrum estimation for comfort noise generation, the method comprising the steps of:
receiving an input noise component of a signal;
approximating a spectrum of the input noise component using an algorithm over a period of time;
detecting an absence of speech signals;
generating comfort noise based on the approximating step when the absence of speech signals is detected; and
performing a variable precision calculation of a least mean square error and at least one least mean square coefficient to make the algorithm substantially independent of variations in noise levels;
wherein the spectrum of the input noise component is substantially constant over the period of time.
21. A system for implementing a spectrum estimation for comfort noise generation, the system comprising:
an encoder adapted to receive an input noise component of a signal for approximating a spectrum of the input noise component using an algorithm over a period of time;
a detector for detecting an absence of speech signals; and
a comfort noise generator for generating comfort noise based on the approximation of the spectrum when the absence of speech signals is detected;
wherein the spectrum of the input noise component is substantially constant over the period of time and
wherein an internal check is performed to ascertain that the input noise component is within approximately 6 dB of a noise floor.
22. The system of claim 21 , wherein the encoder further shapes the input noise to a spectrum of a predicted signal using an inverse predictor.
23. The system of claim 21 , wherein approximating to at least one of noise spikes and speech segments is prevented.
24. The system of claim 21 , wherein the algorithm is a least mean square algorithm.
25. The system of claim 21 , wherein the algorithm is a leaky least mean square algorithm.
26. The system of claim 21 , wherein the algorithm is a normalized least mean square algorithm.
27. The system of claim 21 , wherein the algorithm is a linear predictive coding algorithm.
28. The system of claim 21 , wherein the generated comfort noise is substantially uniform.
29. The system of claim 21 , the algorithm is normalized for making the approximation of the spectrum substantially independent of signal amplitude variations.
30. The system of claim 21 , wherein filter parameter encoding into at least one reflection coefficients is segregated from the approximation of the spectrum for transmitting at least one silence insertion descriptor.
31. The system of claim 30 , wherein interoperability between systems is enhanced.
32. The system of claim 31 , wherein MIPS and memory are efficiently utilized.
33. The system of claim 21 , further comprises a synthesis filter for filtering the input noise component.
34. The system of claim 33 , wherein the synthesis filter is defined as follows:
H
(
z
)
=
1
∑
i
=
0
M
w
i
z
-
i
wherein M represents a number of taps, w represents a predictor coefficient and H is a function of variable z.
35. The system of claim 33 , wherein the synthesis filter is a 10 th order synthesis filter.
36. The system of claim 21 , wherein the encoder further comprises:
a detector for detecting noise between speech data;
an adaptor for adapting to the noise; and
silence insertion descriptor creator for creating silence insertion descriptors based on the adapting step when speech is inactive.
37. The system of claim 36 , wherein silence insertion descriptors are generated by converting at least one direct form coefficients to at least one reflection coefficients as represented by:
A
m
-
1
(
z
)
=
[
A
m
(
z
)
-
B
m
(
z
)
k
m
]
[
1
-
k
m
2
]
m
=
M
-
1
,
M
-
2
,
…
,
1
B m ( z )= z −m A m ( z −1 )
38. The system of claim 36 , wherein silence insertion descriptors are decoded by converting at least one reflection coefficients to direct form coefficients as represented by:
A
0
(
z
)
=
B
0
(
z
)
=
1
A
m
(
z
)
=
A
m
-
1
(
z
)
+
z
-
1
B
m
-
1
(
z
)
k
m
m
=
1
,
2
,
…
,
M
-
1
B
m
(
z
)
=
z
-
m
A
m
(
z
-
1
)
m
=
1
,
2
,
…
,
M
-
1
39. The system of claim 21 , wherein the encoder further comprises:
a detector for detecting noise between speech data;
an adaptor for adapting to the noise; and
a noise generator for generating enhanced noise based on an average spectrum of the input noise component when speech is inactive.
40. A system for implementing a spectrum estimation for comfort noise generation, the system comprising:
an encoder adapted to receive an input noise component of a signal for approximating a spectrum of the input noise component using an algorithm over a period of time;
a detector for detecting an absence of speech signals; and
a comfort noise generator for generating comfort noise based on the approximation of the spectrum when the absence of speech signals is detected;
wherein the spectrum of the input noise component is substantially constant over the period of time and
wherein a variable precision calculation of a least mean square error and at least one least mean square coefficient is performed to make the algorithm substantially independent of variations in noise levels.