10.703GPPGSM Adaptive Multi-Rate Speech Codec (AMR)Project schedule and open issues for AMRTS
6.1 SMG11 mandate and specification development time scale
The initial mandate from SMG#23 (October 1997) for the AMR codec development is given in  and reproduced below:
– SMG11 and SMG2 start narrowband AMR immediately with the target to complete the codec specifications by 12/98 and the systems specifications during 1999.
– The feasibility assessment of wideband will be completed as soon as possible.
– Wideband may be developed with narrowband or, alternatively, be added to AMR as a later option according to the outcome of the wideband feasibility assessment.
– The AMR selection process will allow the submission of wideband candidates. Every proposing party is obliged to describe an evolution path to include wideband.
– The AMR development should aim at one expandable platform.
SMG will review the status and time scale of the program at appropriate intervals.
Initial target dates for the finalized AMR specifications (ready for SMG approval) were given in :
– December 1998: source codec, channel codec, bad frame handling, in-band signaling of codec mode – transmission aspects and definition of parameters, in-band signaling of channel metric and side information – transmission aspects (bit allocation and channel protection)
– December 1998 highly desirable: VAD/DTX/comfort noise generation, definition of channel metric and side information parameters, example of codec mode adaptation, layer 3 signaling
– June 1999: AMR TRAU frames, channel performance tables (GSM 05.05), TFO, test sequences
– December 1999: performance characterization, [minimum performance of adaptation algorithms]
At SMG11#7, the target dates were updated so that most of the AMR specifications could be approved as part of the GSM Release 98 in February 99. The new target dates for the completion of the specifications were set as follows:
February 99: Source codec and channel codec, bad frame handling, in-band signaling of codec mode indication and control and/or quality measurement. VAD/DTX/comfort noise generation, definition of quality reporting, example of codec mode adaptation, layer 3 signaling, AMR TRAU frames, channel performance tables
– March 1999: Performance characterization
– June 1999: TFO, Test sequences, Test specifications
– AMR Wideband mode and Noise Suppresser (tbd)
SMG#28 (2/99) must select the AMR channel coding between two alternatives evaluated by SMG2 and SMG11. Because of this incertitude on the Channel Coding, the definition of the channel performances tables (Update on the 05.05) has been delayed to SMG#29. Likewise, two 16 kbps A-TRAU schemes are under discussion in SMG2. As a result, the final proposal for the 08.60 and 08.61 updates was also delayed to SMG#29.
6.2 Phased Introduction of Capability
The first and only phase of the AMR standard should be completed by February 1999 and approved at SMG#28.
The possibility to start a second phase of the AMR development as described in section 3.0 will be assessed at the end of the first phase of development.
Future optional enhancements to the AMR standard (Wideband mode, Noise suppresser) will be introduced in 1999. The detailed schedule for these enhancements is tbd.
6.3 Development Plan and Status
The development of the AMR speech codec followed a traditional speech codec selection phased approach. Organizations interested in proposing an AMR candidate were first required to present subjective listening test results of their solution according to a previously agreed Qualification Test Plan , in addition to preliminary technical information on their proposal . The best candidates were pre-selected by SMG11 to take part in the selection phase. This pre-selection was approved by SMG#26.
After analysis of the selection test results, the candidate providing the best performances was recommended to SMG#27 to serve as the basis for the production of the AMR standard. After the SMG approval, the development process was completed with the optimization and verification phases. The completion of final verification items and characterization tests and the preparation of the TR on the Characterization of the AMR Speech Codec are scheduled for 1-2Q99. Additional information on these tasks is provided below:
1. Qualification Phase: The five most promising codec solutions were qualified to enter the selection phase after analysis of the in-house listening test results and technical information on the proposed solutions. The Qualification Phase was completed in June 1998. The results were presented and approved at SMG#26 .
2. Selection Phase: Seven independent listening laboratories performed formal subjective listening tests on the five pre-selected candidates to find the optimum codec solution. The best solution (ENS1 proposal) was selected by SMG11#7 in September 1998 and recommended as the preferred AMR choice to SMG. The selection test results were presented and approved at SMG#28  with the recommendation to select the ENS1 proposal. An extract of the selection test report showing the key AMR performances obtained by the best candidates to the selection phase is included in Annex 3. Following the SMG11 recommendation, SMG also agreed that the selection results did not justify to initiate a long optimization phase, but approved a 3 month optimization essentially restricted to the correction of the bugs identified by the ENS consortium members, proponents of the selected solution.
3. Verification Phase: The verification items were distributed in two phases in SMG11 AMR#12 in October 1998. The first phase included the checking of the bit-exactness of the ENS codec with the executable and C-Code delivered earlier in the development process to ETSI. The second phase included verification items related to the codec performance and compatibility with special operational conditions and input signals (DTMF, announcement tones, language dependency, overload, special inputs, influence of VAD/DTX, etc.). The verification items were allocated to SMG11 members (other than the ENS consortium members) in SMG11 AMR#12. The first part of the verification phase was completed by SMG11#9 in January 1999 without any issue being identified. The second part was partially completed by SMG11#9. Because of the two channel coding alternatives and the two VAD options proposed for standardization to SMG#28, few additional verification will be required especially on the operation of the VAD options and generation of the Comfort Noise in DTX mode. These tasks will be completed for SMG11#10 (4/99). SMG11#9 considered that the preliminary verification results were satisfying enough to proceed with the approval of the specifications in their present status.
4. Optimization Phase: Final corrections were brought to the C-code of the selected solution in the 11/98 to 01/99 time-frame. Performance enhancements were essentially selected based on expert listening tests. The optimization phase leading to the final version of the code was completed for SMG11#8 in November 1998. It was restricted to the ENS consortium members. The code was later modified before it was released to SMG11 members participating to the verification phase. A new set of modifications was identified after the review of the verification phase results and the approval of the AMR specifications in SMG2 and SMG11.
In addition, SMG2#28 proposed to evaluate a modified channel coding scheme restricting the convolutional codes to the existing polynomials. Additional simulation and subjective listening tests were required to assess the respective performance of the different channel coding options. After review of the results, SMG2 and SMG11 decided to propose both alternatives to SMG#28 with a request to adopt only one scheme for the AMR standard.
5. Selection of VAD/DTX: As requested in the design constraints, the selection winner proposed a dedicated AMR Voice Activity Detection and Comfort Noise Generation algorithms and a new Discontinuous Transmission scheme. The corresponding specifications were presented in SMG11#8 in November 1998. The inclusion of the corresponding code in the version delivered for the verification phase was also approved by SMG11#8.
The finalization of the AMR VAD/DTX standard was open to contributions from any SMG11 member as defined in . A second VAD alternative was introduced in SMG11#8 and the corresponding specification presented in the Ad Hoc SMG11#8Bis meeting (12/98). After a review of the new scheme, SMG11#9 decided to include this VAD as a second option in the AMR standard with a strong desire to remove one option by SMG#29, after the completion of the remaining verification items.
6. Edition of new Recommendations and Change Requests to new recommendations: The new recommendations and available change requests were approved in January 1999 by SMG2#29 and SMG11#9. They are presented to SMG#28 for approval. Change Requests on 05.05, 08.60, 08.61, 08.62 and the 11 series are expected for approval by SMG#29.
7. Characterization Phase: The performances of the AMR codec will be evaluated in multiple environmental conditions and with a diversity of input signals. The ETSI Technical Report (ETR) on the Performance Characterization of the AMR codec will be prepared for SMG#29 (6/99). The characterization test plan was approved at SMG11#9. The tests are scheduled for 1Q99.
8. Wideband AMR mode development Phase: The selection and possible standardization of a wideband mode could take place after the completion of a dedicated feasibility study. The corresponding report is expected for SMG#29 (6/99).
9. Noise Suppression development Phase: The evaluation and possible selection and standardization of a noise suppresser has been initiated in SMG#27 after the completion of the AMR narrowband selection. The corresponding selection tests are scheduled for mid-99.
10. Future development phases: To be defined
6.4 Road Map for the approval of the AMR standard
SMG#28 – February 8-12, 1999: Approval of AMR as part of the GSM Release 98
SMG11#10 – April 1999: Review of the results of the remaining verification items and characterization tests
SMG2#30 – April 12-16, 1999 Approval of CRs on 05.05, 08.60 & 08.61
SMG11#11 – May 1999 Review and approval of the ETR on the AMR Characterization Performances
Approval of the AMR wideband mode feasibility study report
SMG#29 – June 21-25, 1999: Review and approval of the remaining items of the development process (CRs on 05.05, 08.60&08.61, TFO, Characterization test results, Test sequences and Test specifications)