Upload
barbra-hensley
View
213
Download
1
Tags:
Embed Size (px)
Citation preview
1 1
Title: Proposal for Addition of Fixed Rate 2 kbps Mode to EVRC-B
Abstract: This contribution provides a proposal for adding a fixed rate 2 kbps vocoder to EVRC-B family of vocoders. The new low rate vocoder is intended to be used in xHRPD application.
Source: Rhys W. Robinson ([email protected]).
TerreStar Networks
Sourabh Gupta ([email protected])
DBSD North America (ICO)
Eric Jacks ([email protected])
Skyterra Communications
Alok Gupta ([email protected])
Qualcomm Incorporated
Date: March 01, 2010
Recommendation: Review and adoptNotice
Skyterra Communications, DBSD North America (ICO), TerreStar Networks, and QUALCOMM Incorporated grant a free, irrevocable license to 3GPP2 and its Organizational Partners to incorporate text or other copyrightable material contained in the contribution and any modifications thereof in the creation of 3GPP2 publications; to copyright and sell in Organizational Partner’s name any Organizational Partner’s standards publication even though it may include all or portions of this contribution; and at the Organizational Partner’s sole discretion to permit others to reproduce in whole or in part such contribution or the resulting Organizational Partner’s standards publication. Skyterra Communications, DBSD North America (ICO), TerreStar Networks, and QUALCOMM Incorporated are also willing to grant licenses under such contributor copyrights to third parties on reasonable, non-discriminatory terms and conditions for purpose of practicing an Organizational Partner’s standard which incorporates this contribution.
This document has been prepared by Skyterra Communications, DBSD North America (ICO), TerreStar Networks, and QUALCOMM Incorporated to assist the development of specifications by 3GPP2. It is proposed to the Committee as a basis for discussion and is not to be construed as a binding proposal on Skyterra Communications, DBSD North America (ICO), or QUALCOMM Incorporated. Skyterra Communications, DBSD North America (ICO), TerreStar Networks, and QUALCOMM Incorporated specifically reserve the right to amend or modify the material contained herein and nothing herein shall be construed as conferring or offering licenses or rights with respect to any intellectual property of Skyterra Communications, DBSD North America (ICO), TerreStar Networks, or QUALCOMM Incorporated other than provided in the copyright statement above.
3
Outline
• EVRC-B Frame Type Classifications• 2 kbps Vocoder Choices• 2 kbps Vocoder Approach • Encoder-Decoder Block Diagram• Conclusions
4
EVRC-B Frame Type Classifications
• In EVRC-B Vocoder, a 20 ms speech frame is classified into following frame types: Silence
Eighth-rate (16 bits) vocoder to transmit the background noise Unvoiced
Quarter rate NELP(Noise Excited Linear Prediction) (40 bits) Voiced
Quarter (40 bits) and Full rate (171 bits) PPP (Prototype Pitch Period) or half rate CELP (Codebook Excited Linear Prediction)
Up-transientTransition between voiced and unvoiced segmentsFull/half rate (171 bits) CELP
Transient Usually in the middle or towards the end of the voiced segment, where pitch is
changingFull/half rate (171 bits) CELP
Down-transient Transition from voiced to unvoiced segmentHalf rate (80 bits) CELP
5
2 kbps Vocoder Choices
• Not many choices! • Quarter rate NELP for unvoiced and may be for down
transient frames • Only quarter rate PPP available for up-transient, voiced
and transient frames • Can work with some changes, like non-differential
encoding of pitch and prototype waveform parameters
6
2 kbps Vocoder Choices (cont.)
• QPPP is ideally suited for purely voiced frame Zero bits on the phase of the prototype waveform
• QPPP can cope with gradual change of pitch over a voiced segment Cannot handle large pitch change
• Use of QPPP for coding all the transient frames result in poor quality
7
2 kbps Vocoder Approach
• Need a new coding scheme to encode the transient frames at 2 kbps
• The other areas that need to be worked or improved in order to have a good quality 2 kbps vocoder Pitch estimation: Re-visit range and accuracy Revisit QPPP to make it more robust in the absence of FPPP Fine tuning of mode decision algorithm
8
Top Level Changes
• FCELP, HCELP or FPPP not used• New scheme for coding transient frames at 2 kbps• Enhanced pitch estimation • QPPP for voiced frame with modifications like non-
differential quantization• Mode-decision tuning
9
Decoder Block Diagram
FER rate-estimation
text
Quarter-rate PPPExcitation
LSP Dequantize
Synthesis Filter
Post Filter
output speech
Quarter-rate TransientExcitation
LSP Dequantize
Synthesis Filter
Post Filter
output speech
LSP Dequantize
Dequantize Gain
random noise
XSynthesis
FilterPost Filter
output speech
Excitation
Eighth-rate
Quarter rate PPP
received packet
Quarter rate Transient
LSP Dequantize
Dequantize Gain
random noise
XSynthesis
FilterPost Filter
output speech
ExcitationShaping
Filter
Quarter-rate NELP
10
Encoder Block Diagram
Pitch
Speech
Mode
Mode
QLSP
QLSP
LSP
Mode
No
No
No
Yes
Yes
Yes
Pre-Processing (Noise Suppression,
High-pass Filter)
LPC Analysis(lpcanaly
s)
Get Residual(getresidual)
Voice Activity
Decision(select_r
ate)Open-Loop Pitch Calculation
(fndppf, get_nacf_at_pitch)
Mode Decision(new_mode_de
cision)
Unvoiced
Voiced
Silence
1/8 Rate LSP VQ
(enc_lsp_vq_8)
1/4 Rate LSP VQ
(enc_lsp_vq_16)
1/8 RateGain-shaped
Random Noise(GetExc800bp
s)
1/4 RateGain-shaped
Random Noise(polezero_filter, quantize_uvg)
Update Memories(interpol, lsp2a,
SynthesisFilter, weight, fir, iir, ZeroInput)
Update Memories
(interpol, lsp2a, SynthesisFilter, weight,
fir, iir, ZeroInput)
Get Residual(interpol,
lsp2a, getresidual)
Get Residual(interpol, lsp2a,
getresidual)
Extract Prototype
PitchPeriod
Fourier Series
Restore Past Prototype Fourier Series
No Signal Modification
(RCELP)
No Signal Modification
(RCELP)
1/4 Rate Voiced
LSP VQ
Get ResidualSignal
Modification (RCELP)
1/4 Rate Amplitude
QuantizationWI Synthesis
Update Memories
Transient Coding
Update Memories
(ZeroInput)
Get Residual(interpol, lsp2a,
getresidual)
Signal Modification
(RCELP)(Interpol_delay, ComputeACB)
LPC to LSP(a2lsp)
A
A
A
1/4 Rate Voiced
LSP VQ
A
Green: unchanged blocksBlue: modified blocksYellow: new block