You are on page 1of 18

1

Analysis of Echo
Cancellation Techniques
in Multi-Perspective Smart
Classroom
RAMESH GUNTHA
AMRITA CENTER FOR WIRELESS NETWORKS AND
APPLICATIONS , AMRITA UNIVERSITY
RAMESHG@AM.AMRITA.EDU
Outline 2

 Motivation
 Background information/Related works
 Proposed Method
 Results
 Summary
 Future Work
 References
Motivation 3
 Echo cancellation technology plays critical role in
ensuring high quality audio communication
 The current echo cancellation technology works
only when the audio goes in and out of a single
computer in a given room
 Our Smart classroom e-Learning system requires 3
computers for a given classroom to capture and
stream HD video from 5 video cameras placed at
different angles or perspectives
 It is done to achieve gaze alignment across all the
remote participants by showing the appropriate
perspective at each remote classroom display
based on the current teaching mode of either
lecturing or interaction
 Since the audio is processed through three
computers in a given room, we had to augment
traditional echo cancellation technology with new
techniques to achieve echo cancellation
Motivation 4

 Smart Classroom Setup with Teacher’s, Local


Students’ and two remote student’s views
 Gaze alignment: All participants at all locations are
watching the teacher
Background information/Related works
 High quality audio communication is one of the most 5
important factors of a successful live interactive e-Learning
system, without echo cancellation the audio
communications suffer greatly.
 Echo cancellation technology works by sampling and
analyzing the audio signals being played through speaker
and then comparing with incoming audio signals from
microphone and removing the matched audio
Background information/Related works 6
 Current Echo Cancellation techniques
 A lot of research has been done to improve the multi-
channel audio quality in live interactive systems by
eliminating acoustic echo and double-talk by using..
 Least mean squares (LMS)
 Normalized LMS (NLMS)
 Proportionate NLMS (PNLMS),
 µ-law PNLMS (MPNLMS),
 Vector-space-based adaptive filtering algorithm.
 Now the echo cancellation is a standard feature in
audio codecs like SPEEX
 The echo cancellation techniques require all the
audio played and received in a given room to pass
through the single instance of echo cancellation
implementation i.e. a computer
Background information/Related works
Our smart classroom has three computers with echo 7
cancellation technology implemented, each computer's
speaker playing audio .from a different remote location
resulting in an elaborate echo path as shown in picture below
Proposed Method - 3 Techniques 8
 We implemented 3 different
techniques by which we
solved the echo problem
 Central Audio
 Mode-Talk
 Hand-Held-Talk.
Proposed Method – Central Audio
9

 The audio of all the remote nodes is played through the


central computer's display
 The Problem is reduced to single computer problem, by
which the existing techniques are able to cancel the
echo successfully
Proposed Method – Central Audio
10

 The audio of all the remote nodes is played through the


central computer's display
 Audio is transmitted separately to the Video
 The Problem is reduced to single computer problem, by
which the existing techniques are able to cancel the echo
successfully
Proposed Method – Mode-Talk
11
 Modes
 Lecture – Only teacher is
speaking and others are
listening
 Interaction with a
Student – Teacher and
the respective student
are speaking and others
are listening

 The microphones of non active speakers are muted


 The Problem is reduced to single computer problem, by
which the existing techniques are able to cancel the
echo successfully
 Directional audio maintained
Proposed Method – Hand-Held-Talk
12
 Hand-Held Mike
 Is very low sensitivity
 Captures only the
speaker’s voice
 Does not capture the
remote classroom’s
voice coming through
the displays/speakers

 Echo does not occur due to the very low sensitivity of the
Mike
Results/Summary 13
Evaluation of Echo cancellation techniques

Criteria of Smart classroom


Echo factors
cancellation Directi
Hands-free Free-
Technique onal
microphone talk
audio

Central Audio Yes No Yes

Mode-Talk Yes Yes No

Hand-Held-Talk No Yes Yes


Results/Summary 14
User’s feedback – Central Audio
Echo
cancellation Rating
Suggested
Technique Feedback (Number of
Modes
users liked)

Users did not like this technique as it


affects gaze alignment as the
sound for all the remote nodes is
played through the central
display's speakers.

Central Causes confusion as to where to


None 1/5
Audio look during the initial part of the
conversation and hence results in
loss of immersion.

Since audio and video are


transmitted as separate streams it
could cause lip-out-of-sync issue.
Results/Summary 15
User’s feedback – Mode Talk
Echo
cancellation Rating
Suggested
Technique Feedback (Number of
Modes
users liked)

Users liked this technique during


4/5 (For
lecturing and interaction modes
Lecturing
but not during meeting and
and
discussion modes
Interaction)
Lecturing
Mode-Talk
Users felt little inconvenience as Interaction
2/5 (For
they could not express their views
Discussion
spontaneously and had to wait for
and
teacher's mediation and
Meetings)
permission
Results/Summary 16
User’s feedback – Hand-Held-Talk
Echo
cancellation Rating
Suggested
Technique Feedback (Number of
Modes
users liked)

Users are ok with this technique


during all the teaching modes, but
not greatly liked it as they have to
wait for the microphone.

Lot of times it's observed that users Lecturing


Hand-Held- get immersed in the session and Interaction
3/5
Talk start to express their views Discussion
spontaneously but then realize that Meetings
they do not have the microphone
and have to wait till the
microphone arrives, this often
causes the entire session to loose
immersive feeling.
Conclusions/Future Work 17

 None of the techniques pass against all the


criteria and hence do not provide 100%
satisfaction to the users during all the teaching
modes. Hence we have to choose a certain
technique during a certain type of teaching
mode or session.
 The system should be tested with more than 3
classrooms to see the scalability of the system
 Further studies and related experiments are
needed to come up with more innovative
techniques which work well in all the modes of
teaching.
References 18
 Narayanankutty, R., Rangan, P. V., Gopalakrishnan, U., Krishnan, R., Guntha. R., Hariharan, B. (2015). Automatic multi-perspective switching
for gaze alignment in e-Learning systems. IJAER, Volume 10, Number 16, pp 37303-37310, 2015.

 Rangan, P. V., Hariharan, B., Krishnan, R., Narayanankutty, R., Gopakumar, S. A., & Gopalakrishnan, U. (2014). U.S. Patent Application No.
14/216,083.

 Acker, S. R., & Levitt, S. R. (1987). Designing videoconference facilities for improved eye contact. Journal of Broadcasting & Electronic
Media, 31(2), 181-191.

 Ho, Y. S., & Jang, W. S. (2015, July). Gaze correction using 3D video processing for videoconferencing. In Signal and Information Processing
(ChinaSIP), 2015 IEEE China Summit and International Conference on (pp. 496-499). IEEE.

 Ko, E., Jang, W. S., & Ho, Y. S. (2015). Eye Gaze Correction for Video Conferencing Using Kinect v2. In Advances in Multimedia Information
Processing--PCM 2015 (pp. 571-578). Springer International Publishing.

 Pan, Y., Oyekoya, O., & Steed, A. (2015). A surround video capture and presentation system for preservation of eye-gaze in
teleconferencing applications. Presence, 24(1), 24-43.

 Sondhi, M. Mohan, Dennis R. Morgan, and Joseph L. Hall. "Stereophonic acoustic echo cancellation-an overview of the fundamental
problem." Signal Processing Letters, IEEE 2.8 (1995): 148-151.

 Gay, Steven L. "An efficient, fast converging adaptive filter for network echo cancellation." Signals, Systems & Computers, 1998.
Conference Record of the Thirty-Second Asilomar Conference on. Vol. 1. IEEE, 1998.

 Gansler, Tommas, et al. "Double-talk robust fast converging algorithms for network echo cancellation." Speech and Audio Processing, IEEE
Transactions on 8.6 (2000): 656-663.

 Benesty, Jacob, Dennis R. Morgan, and Man Mohan Sondhi. "A better understanding and an improved solution to the specific problems of
stereophonic acoustic echo cancellation." Speech and Audio Processing, IEEE Transactions on 6.2 (1998): 156-165.

 Deng, Hongyang, and Milos Doroslovacki. "Proportionate adaptive algorithms for network echo cancellation." Signal Processing, IEEE
Transactions on 54.5 (2006): 1794-1803.

 Shilpa, B., Balaji Hariharan, and G. Uma. "Echo cancellati on in a virtual classroom environment." Multimedia and Broadcasting
(APMediaCast), 2015 Asia Pacific Conference on. IEEE, 2015.

 Tsao, Yu, Shih-Hau Fang, and Yao Shiao. "Acoustic echo cancellation using a vector-space-based adaptive filtering algorithm." IEEE Signal
Processing Letters 22.3 (2015): 351-355

You might also like