System and method for acoustic echo cancellation

Patent metadata

Field	Value
Publication number	US-9628141-B2
Application number	US-201314060141-A
Country	US
Kind code	B2
Filing date	Oct 22, 2013
Priority date	Oct 23, 2012
Publication date	Apr 18, 2017
Grant date	Apr 18, 2017

Abstract

A system and method are presented for acoustic echo cancellation. The echo canceller performs reduction of acoustic and hybrid echoes which may arise in a situation such as a long-distance conference call with multiple speakers in varying environments, for example. Echo cancellation, in at least one embodiment, may be based on similarity measurement, statistical determination of echo cancellation parameters from historical values, frequency domain operation, double talk detection, packet loss detection, signal detection, and noise subtraction.

First claim

The invention claimed is:

1. A system for cancellation of acoustic echo comprising:
a. a source of audio input, wherein the source of audio input generates an audio signal;
b. a network, wherein the network is connected to the source and to an echo cancellation module, wherein the network transmits said audio signal from the source to the echo cancellation module; and
c. the echo cancellation module, wherein the module:
i. converts the audio signal into a frequency domain,
ii. performs similarity measure on the converted audio signal,
iii. performs delay estimation mathematically using an expression for similarity measure on the converted audio signal, wherein the expression for similarity measure comprises the mathematical expression:

Diff
⁡

(
i
)

=

∑

m
=
1

m
=
K

⁢

⁢



NearEnd
⁡

(
m
)

-

FarEnd
⁡

(

m
+
i

)



⁢

⁢
with

⁢

i
=
1

,
…
⁢

,

N
-
K

wherein i represents a variable index from 1 to (N−K), m represents a summation index, N represents frames in a near-end speech signal, and K represents frames in a far end speech signal,

iv. performs echo parameter estimation for a validation model,
v. performs statistical echo validation on the converted audio signal using the validation model,
vi. detects speech in the converted audio signal, and
vii. detects double-talk in the converted audio signal.

2. The system of claim 1 , wherein said source of audio input comprises a receiver.

3. The system of claim 1 , wherein the echo cancellation module applies Fast Fourier Transform to convert the audio signal into the frequency domain.

4. The system of claim 1 , wherein performing similarity measure further comprises one or more of:
a. transforming signals into the frequency domain, wherein the transforming comprises Fast Fourier Transform;
b. normalizing spectra; and
c. performing band pass filtering.

5. The Fast Fourier Transform of claim 4 , wherein said Fast Fourier Transform operates using 128 bins.

6. The system of claim 1 , wherein performing delay estimation comprises examining and comparing recent frames from a first signal and a second signal.

7. The system of claim 6 , wherein said first signal comprises a near-end signal and said second signal comprises a far-end signal.

8. The system of claim 1 , wherein performing echo parameter estimation comprises utilizing a histogram.

9. The system of claim 1 , wherein detecting speech comprises variability based at least in part on a spectrum of consecutive frames and an estimated signal power.

10. The system of claim 1 , wherein detecting double talk further comprises:
a. controlling an amount of echo removed when speech is present;
b. determining a presence of far-end and near-end speech; and
c. analyzing similarity measure.

11. A method for acoustic echo cancellation comprising the steps of:
a. initializing echo model parameters;
b. analyzing an audio signal for speech;
c. determining if speech has been detected in said audio signal, wherein if speech has not been detected in said audio signal, continuing to analyze said audio for speech;
d. estimating echo delay in said audio signal and validating an echo model for said audio signal if speech has been detected in said audio signal, wherein the estimating echo delay and validating an echo model further comprises:
i. measuring and summing a distance for each frame of the audio;
ii. estimating the echo delay mathematically using the mathematical expression for similarity measure, wherein the mathematical expression comprises:

Diff
⁡

(
i
)

=

∑

m
=
1

m
=
K

⁢

⁢



NearEnd
⁡

(
m
)

-

FarEnd
⁡

(

m
+
i

)



⁢

⁢
with

i
=
1

,
…
⁢

,

N
-
K

Assignees

• Interactive Intelligence Inc
• Interactive Intelligence Group Inc

Inventors

• Wyss Felix Immanuel
• Vergin Rivarol
• Iyer Ananth Nagaraja
• Ganapathiraju Aravind
• Vlack Kevin Charles
• Cheluvaraja Srinath

Classifications

• H04B3/23

  using a replica of transmitted signal in the time domain, e.g. echo cancellers · CPC title

• G10L2021/02082

  the noise being echo, reverberation of the speech · CPC title

• H04M9/082

  using echo cancellers (echo cancellers per se H04B3/23) · CPC title

• H04B3/234 Primary

  using double talk detection · CPC title

• G10L21/0208

  Noise filtering · CPC title