10.703GPPGSM Adaptive Multi-Rate Speech Codec (AMR)Project schedule and open issues for AMRTS
The complexity requirements are valid for all phases of the AMR development and are separate for channel coding, speech coding and DTX algorithms.
Channel coding including control loop management algorithms:
A. wMOPS(AMR-HR ch. codec) £ 3.0 wMOPS » 1.1 ´ wMOPS(HR ch. codec)
B. wMOPS(AMR-FR ch. codec) £ 5.7 wMOPS » 2.1 ´ wMOPS(HR ch. codec)
C. RAM(AMR-HR ch. Codec) £ 3.5 kwords » 1.1 ´ RAM(HR ch. Codec)
D. RAM(AMR-FR ch. Codec) £ 6.6 kwords » 2.1 ´ RAM(HR ch. Codec)
E. ROM(AMR ch. Codec) £ 5.0 kwords » 5.6 ´ ROM(HR ch. codec)
F. FOM(AMR HR ch. Codec) £ 4.1 » 1.0 ´ FOM(HR ch. codec)
G. FOM(AMR FR ch. Codec) £ 8.2 » 2.0 ´ FOM(HR ch. codec)
H. Program ROM(AMR ch. Codec.) £ 4 ´ Program ROM(HR channel. Codec.)
Speech coding (excluding VAD/DTX):
I. wMOPS(AMR speech codec) £ 24 wMOPS » 8.1 ´ wMOPS(FR speech codec)
J. RAM(AMR speech codec) £ 10 kwords » 2.1 ´ RAM(EFR speech codec)
K. ROM(AMR speech codec) £ 17 kwords » 3.2 ´ ROM(EFR speech codec)
L. FOM(AMR speech codec) £ 27.5 = 8.0 ´ FOM(FR speech codec)
» 1.6 ´ FOM(EFR speech codec)
M. Program ROM(AMR sp. Codec) £ 3 ´ Program ROM(EFR speech. Codec.)
Additional complexity for VAD/DTX operation:
N. wMOPS(AMR VAD/DTX) £ 0.2 wMOPS » 2 ´ wMOPS(EFR VAD/DTX)
O. RAM(AMR VAD/DTX) £ 0.6 kwords » 0.1 ´ RAM(AMR speech codec)
P. ROM(AMR VAD/DTX) £ 0.5 kwords » 4 ´ ROM(EFR VAD/DTX)
FOM = Figure of merit, as described in doc. AMR-9
Program ROM is computed as the number of basic instructions
The control loop management algorithms are intended to include all the additional algorithms beyond speech and channel codec that are needed for codec mode adaptation: channel metric estimation, adaptation algorithm, coding and decoding of the inband signaling.
Complexity calculation rules:
The same complexity evaluation methodology as used in the past for GSM HR and EFR codecs (based on ETSI fixed-point basic operations) will be used for complexity evaluation of the AMR codec. Detailed procedure for each phase is the following:
Qualification: Complexity evaluation may be based on floating point code. The results should nevertheless be presented as ETSI FOM, wMOPS, and memory figures even though they are allowed to be estimated from a floating point code. Requirements shall be checked according to the assessment methodology given in document AMR-9 (Complexity and delay assessment).
Selection: ETSI methodology based on fixed point code (Basic op. Counters, i.e. Worst observed case)
Verification/characterization: ETSI methodology based on fixed point code (Theoretical worst case)
Arithmetic used in codec proposals:
Qualification: Fixed point or floating point code
Selection: Fixed point code (using ETSI set of basic operations)
Verification/characterization: Fixed point code (using ETSI set of basic operations).
A few new arithmetic operators could be added to the ETSI set of basic operators but only to be applied after selection. The detailed definition of these operators is left to the complexity subgroup of AMR. Deadline for provision is 4 weeks after selection.
At least one codec mode at HR should be consistent with 8 kbit/s sub multiplexing on the A-ter interface. This implies the constraint of providing at least one codec mode in HR mode operating at a source codec bit-rate below 6.85 kbit/s. No constraint is applied on the upper limit of the source codec bit-rate in HR mode.
The AMR codec shall support A-ter sub-multiplexing at 16 kbit/s.
The maximum source coding rate for FR channel modes is limited by the constraint of the maximum allowed rate to achieve 16 kbit/s submultiplexing. This implies a maximum source coding rate for FR codec modes of 14.4 kbit/s.
These figures do not take into account the probable need of in-band signaling for TFO operation.
Codec mode signaling: transmitted in-band on the radio interface
Channel measurement signaling: transmitted in-band on the radio interface
Signaling can be different on the up and down links.
Adaptation may operate independently on the up- and down-links.
The working assumption should be not to use the stealing flags for in-band signaling. This assumption can be reverted only it is proved that the stealing flags may be used without degrading the FACCH detection performance as specified in 05.05. Decision to revert the working assumption has to be taken by SMG2.
In the static conditions of qualification tests, four codec modes (for each of the two channel modes) are tested although the codec candidates may include more modes. The codec proponent may select up to 4 codec modes to the qualification test as he wishes, but the chosen codec modes must be the same throughout all the static tests. Also the same codec modes shall be used in the dynamic tests.
Channel mode handovers will be executed in the same way as existing intra-cell handovers. However, the algorithm used to determine when and whether to perform an AMR handover will be new and specific to the BSS manufacturer.
Channel mode signaling: transmitted out of band on the radio interface
The up- and downlink(of the same air-interface) shall use the same channel mode.
Channel mode control is located in the network.
Tandem Free Operation (TFO):
The AMR codec shall support Tandem Free Operation. TFO shall not decrease the capacity gain achievable using the AMR codec.
TFO mode can be operated only if both terminals (e.g. up-link MS-A to network and downlink, network to MS-B) use the same speech coding bit-rate and algorithm. (In principle this does not imply that the same channel mode has to be used, but TFO can be achieved only if the AMR coder is capable of operating at the same source rates for both FR and HR channel modes.)
It is not mandatory to have a common speech codec rate in FR and HR modes.
In order to support TFO, AMR must be designed such that all HR codec contained in the proposal can be supported over a full rate channel. The in band signaling in full rate channel must be designed such that it is possible to signal the operation of all HR and FR speech codecs.
It is possible to use an escape mechanism between the FR and HR mode and enter a new mode of operation of the in band signaling to switch between the HR codec modes in the full rate channel. The escape mechanism is only allowed when entering or leaving the TFO operation mode, and not while operating in TFO.
The channel coding and in band signaling complexity corresponding to the support of all HR codecs in a full rate channel must be included in the FR channel coding complexity figures.
The channel coding required for the support of the HR codec in the full rate channel should be defined as part of the proposal. The corresponding code does not need to be included in the executable to be delivered prior the selection tests. It must be included in the C-Code to be delivered for the evaluation of the channel coding complexity.
Discontinuous Transmission (DTX):
The AMR codec shall support DTX operation. The increase in radio channel activity in terms of average transmission power during speech inactivity shall not significantly affect the gain of DTX operation i.e. the interference reduction and the battery saving should be similar to that of current DTX operation. The VAD algorithm and Comfort Noise generation algorithm are part of the specification of the AMR codec.
To limit the impact of in-band signaling on the DTX gain, the increase of the radio channel activity in terms of average transmission power should not exceed 8 % compared to today’s GSM codecs and assuming the same VAD activity.
Under the assumption the TX power could not be modified for DTX operation, the signaling needed during speech pauses for DTX operation as well as codec mode adaptation shall use no more than 16 burst in FR and 12 burst in HR per multiframe.
Since VAD/DTX is developed only after the selection phase, the codec proponents shall commit to provide it.
The qualification rules will not apply to the test conditions with reversed link in DTX mode.
The assumption of reducing the TX power of in-band signaling frames will not be considered until a confirmation of feasibility is received from SMG2-WPB.
The AMR codec and its control will operate without any changes to the air-interface channel multiplexing, with the possible exception of the interleave depth. Conventional TCH-F and TCH-H channels will be used for FR and HR channel modes of the codec.
It shall be possible to operate power control independently of the AMR adaptation. However, operators may choose to optimize the AMR control according to the power control settings. Fast power control may also be introduced provided that the measurement reports are transmitted in-band for AMR codec adaptation control.
Codec mode control:
Codec mode control relating to capacity or radio link quality should be located in the network (BSS)
MS can autonomously select the codec mode of the up-link on the basis of speech source content (e.g. background noise); however, the network should have the option to e.g. override the MS preferred selection or restrict the range of selectable modes.
It shall be possible to upgrade the adaptation algorithm from the network.
Active noise suppression in the qualification/selection phase:
In order to compare all solutions in the same conditions, and select the candidate with the best intrinsic quality, it was decided that noise suppressers would not be included during the qualification and selection phases, or that any noise suppresser integrated to a source codec should be turned off for these tests. The selection and possible standardization of a noise suppresser will then be addressed in a separate phase
The algorithmic round trip delay for HR codec modes shall be less than or equal to the algorithmic round trip delay of the HR GSM codec increased by 10 ms. The algorithmic round trip delay for FR codec modes shall be less than or equal to the algorithmic round trip delay of the EFR GSM codec increased by 10 ms. This has to be ensured in all operating conditions including codec mode switching.
The transmission delay shall be evaluated according to the rules described in doc. AMR-9.
Error concealment techniques of AMR codec candidates shall only rely on soft-output information from the equalizer. This does not preclude any future exploitation of other radio channel parameters in the final AMR system.
The frame size is constrained to be one of the possible values: 5ms, 10ms or 20 ms.
In addition to the signaling needed to support TFO, the AMR in-band signaling shall be expandable (prepared) to signal the use of future AMR modes including signaling the use of the existing GSM FR, GSM HR and GSM EFR speech coders, one or two wideband modes.
Switching between AMR codecs and extended codecs shall be possible although not tested in qualification and selection phases and the constraint relates to the switching capabilities and not to the quality performance.