Folie 1

1
Graceful Degradation for Digital Radio Mondiale
(DRM) by Ferenc Krämer Gerald Schuller
2
Contents
Contents

• 1. Introduction
• 2. Graceful Degradation approach for DRM
• 3. Results Evaluation
• 4. Conclusions Prospects

3
1. Introduction
Graceful Degradation

• Definition
• System continues operating in case of errors
• Gradual decline in service quality
• Advantage
• No abrupt breakdown
• Allow user to decide on acceptability of service
  quality

4
1. Introduction
Motivation for Graceful Degradation

• Analog broadcasting
• Inherent graceful degradation
• Digital broadcasting
• Lack of gradual degradation in service quality
• Problem
• Frequent loss of service (dropouts)

5
1. Introduction
Goals

• Graceful Degradation for Digital Radio Mondiale
  (DRM)
• Fit seamlessly into DRM standard
• Graceful degradation comparable to classic analog
  AM radio
• Minimum impact on audio quality in undisturbed
  case

6
1. Introduction
Digital Radio Mondiale (DRM)

• Dates
• DRM consortium funded in 1998
• Regular DRM service since 2003
• Characteristics
• Migrate analog AM radio into the digital era
• Increased audio quality (HE-AAC 20kBit/s) with
  higher audio bandwidth and stereo
• Ease of use supplementary services

7
1. Introduction
Reception monitoring
90 _at_ 13s

• 125h monitored (DW, BBC, BR)

• Availability 73.65

8
1. Introduction
Previous approaches

• Source coding
• Comfort noise
• Concealment
• Channel coding modulation
• Unequal error protection
• Hierarchical modulation
• Joint source-channel coding modulation
• Base/enhancement layer

9
2. Graceful Degradation for DRM
Contents

• 1. Introduction
• 2. Graceful Degradation approach for DRM
• 3. Results Evaluation
• 4. Conclusions Prospects

10
2. Graceful Degradation for DRM
Conceptual design
delay
11
2. Graceful Degradation for DRM
Dropout filling
12
2. Graceful Degradation for DRM
Proposed basic version
Transmitter
Receiver
13
2. Graceful Degradation for DRM
Proposed extended version
Transmitter
Receiver
14
2. Graceful Degradation for DRM
Sinusoidal music codec
15
3. Evaluation
Contents

• 1. Introduction
• 2. Graceful Degradation approach for DRM
• 3. Results Evaluation
• 4. Conclusions Prospects

16
3. Evaluation
Listening Test

• MUlti Stimulus with Hidden Reference and Anchors
  (MUSHRA)
• 3 test sessions (speech, music, mixed)
• Material 3 items (MPEG, SQAM), 4 broadcasts
• Dropouts min 1s, max 14s
• 18 test listeners

17
3. Evaluation
Listening Test speech
Average and 95 Confidence Intervals
18
3. Evaluation
Listening Test music
Average and 95 Confidence Intervals
19
3. Evaluation
Listening Test Opinions

• Speech
• Listeners favored the filling of dropouts
• HVXC outperforms all other codecs
• Music
• For 2 out of the 3 music items the sinusoidal
  coder scored significantly better than the speech
  coder
• Sinusoidal codec sounds like a barrel-organ

20
Listening examples
BBCworld - News service

• Dropout anchor muted
• Filled dropouts - HVXC
• Filled dropouts - SIN

seconds
21
Listening examples
sc01 Trumpet, Haydn

• Dropout anchor muted
• Filled dropouts - HVXC
• Filled dropouts - SIN

seconds
22
Listening examples
DW Talk (dropout)

• Dropout anchor muted
• Filled dropouts - HVXC
• Filled dropouts - SIN

seconds
23
4. Conclusions Prospects
Contents

• 1. Introduction
• 2. Graceful Degradation approach for DRM
• 3. Results Evaluation
• 4. Conclusions Prospects

24
4. Conclusions Prospects
Conclusions

• HVXC desirable for speech signals in broadcasting
• Music/speech switching preferable when program
  contains music
• Proposed system easy to implement
• Dropout-filling only dependent on buffer length,
  works over wide range of SNR
• Increase in mean undisturbed audio reception from
  39s to 57s, when buffering 15s.

25
4. Conclusions Prospects
Prospects

• Proposed system
• Implement and test music/speech switching
• Stronger protection of parallel backup stream
  desired
• Improve tracking algorithm for sinusoidal codec
• Further (non-compatible) approaches
• Joint source-channel coding and modulation
• Using multiple descriptions together with
  hierarchical structures

26

• Thank you for your attention!
• Ferenc Krämer
• kraefc_at_idmt.fraunhofer.de
• Gerald Schuller
• shl_at_idmt.fhg.de