6 Speech codec test sequences

06.543GPPTest sequences for the GSM Enhanced Full Rate (EFR)TS

This clause describes the test sequences designed to exercise the GSM enhanced full rate speech transcoder (GSM 06.60 [2]).

6.1 Codec configuration

The speech encoder shall be configured to operate in the non-DTX mode. The VAD and SP flags shall be set to 1 at the speech encoder output.

6.2 Speech codec test sequences

Table 5 lists the location and size of the speech codec test sequences.

6.2.1 Speech encoder test sequences

Twenty-one encoder input sequences are provided. Note that for the input sequences TEST0.INP to TEST3.INP, the amplitude figures are given in 13-bit precision. The active speech levels are given in dBov.

– TEST0.INP – Synthetic harmonic signal. The pitch delay varies slowly from 18 to 143,5 samples. The minimum and maximum amplitudes are -997 and +971.

– TEST1.INP – Synthetic harmonic signal. The pitch delay varies slowly from 144 down to 18,5 samples. Amplitudes at saturation point -4 096 and +4 095.

– TEST2.INP – Sinusoidal sweep varying from 150 Hz to 3 400 Hz. Amplitudes 1 250.

– TEST3.INP – Sinusoidal sweep varying from 150 Hz to 3 400 Hz. Amplitudes 4 000.

– TEST4.INP – Female speech, active speech level: -19,4 dBov, flat frequency response.

– TEST5.INP – Male speech, active speech level: -18,7 dBov, flat frequency response.

– TEST6.INP – Female speech, ambient noise, active speech level: -35,0 dBov, flat frequency response.

– TEST7.INP – Female speech, ambient noise, active speech level: -25,0 dBov, flat frequency response.

– TEST8.INP – Female speech, ambient noise, active speech level: -15,6 dBov, flat frequency response.

– TEST9.INP – Female speech, car noise, active speech level: -35,5 dBov, flat frequency response.

– TEST10.INP – Female speech, car noise, active speech level: -26,1 dBov, flat frequency response.

– TEST11.INP – Female speech, car noise, active speech level: -15,8 dBov, flat frequency response.

– TEST12.INP – Male speech, ambient noise, active speech level: -34,9 dBov, flat frequency response.

– TEST13.INP – Male speech, ambient noise, active speech level: -24,8 dBov, flat frequency response.

– TEST14.INP – Male speech, ambient noise, active speech level: -15,0 dBov, flat frequency response.

– TEST15.INP – Male speech, babble noise, active speech level: -34,1 dBov, flat frequency response.

– TEST16.INP – Male speech, babble noise, active speech level: -24,3 dBov, flat frequency response.

– TEST17.INP – Male speech, babble noise, active speech level: -14,4 dBov, flat frequency response.

– TEST18.INP – Female speech, ambient noise, active speech level: -26,0 dBov, modified IRS frequency response, with many zero frames.

– TEST19.INP – Male speech, ambient noise, active speech level: -36,0 dBov, modified IRS frequency response, with many zero frames.

– TEST20.INP – Sequence for exercising the LPC vector quantisation codebooks and ROM tables of the codec.

The TEST0.INP and TEST1.INP sequences were designed to test the pitch lag of the GSM enhanced full rate speech encoder. In a correct implementation, the resulting speech encoder output parameters shall be identical to those specified in the TEST0.COD and TEST1.COD sequences, respectively.

The TEST2.INP and TEST3.INP sequences are particularly suited for testing the LPC analysis, as well as for finding saturation problems. In a correct implementation, the resulting speech encoder output parameters shall be identical to those specified in the TEST2.COD and TEST3.COD sequences, respectively.

The TEST4.INP and TEST5.INP sequences contain a lot of low-frequency components. In a correct implementation, the resulting speech encoder output parameters shall be identical to those specified in the TEST4.COD and TEST5.COD sequences, respectively.

The TEST18.INP and TEST19.INP sequences contain some “all zeros” frames (silence) in between segments of speech. In a correct implementation, the resulting speech encoder output parameters shall be identical to those specified in the TEST18.COD and TEST19.COD sequences, respectively.

The TEST20.INP sequence was designed to force the encoder to select each of the LPC code indices and each but one of the the ROM table indices of the codec.

The remaining sequences (TEST6.INP to TEST17.INP) were selected on the basis of bringing various input characteristics (background noise) and levels to the test sequence set. In a correct implementation, the resulting speech encoder output parameters shall be identical to those specified in the TEST6.COD to TEST17.COD sequences, respectively.

6.2.2 Speech decoder test sequences

Twenty-one speech decoder input sequences TESTXX.DEC (XX = 0..20) are provided. These are derived from the corresponding TESTXX.INP sequences. In a correct implementation, the resulting speech decoder output shall be identical to the corresponding TESTXX.OUT sequences.

6.2.3 Codec homing sequence

In addition to the test sequences described above, two homing sequences are provided to assist in codec testing. TEST21.INP contains one encoder-homing-frame. TEST21.DEC contains one decoder-homing-frame. The use of these sequences is described in GSM 06.51 [8].

Table 5: Location and size of speech codec test sequences

Disk No.

File Name

No. of frames

Size (bytes)

1/8

1/8

1/8

1/8

TEST0.INP

TEST0.COD

TEST0.DEC

TEST0.OUT

285

91 200

140 220

140 790

91 200

1/8

1/8

1/8

1/8

TEST1.INP

TEST1.COD

TEST1.DEC

TEST1.OUT

285

91 200

140 220

140 790

91 200

1/8

1/8

1/8

1/8

TEST2.INP

TEST2.COD

TEST2.DEC

TEST2.OUT

402

128 640

197 784

198 588

128 640

1/8

1/8

1/8

1/8

TEST3.INP

TEST3.COD

TEST3.DEC

TEST3.OUT

402

128 640

197 784

198 588

128 640

1/8

1/8

1/8

1/8

TEST4.INP

TEST4.COD

TEST4.DEC

TEST4.OUT

301

96 320

148 092

148 694

96 320

1/8

1/8

1/8

1/8

TEST5.INP

TEST5.COD

TEST5.DEC

TEST5.OUT

224

71 680

110 208

110 656

71 680

1/8

1/8

1/8

1/8

TEST6.INP

TEST6.COD

TEST6.DEC

TEST6.OUT

335

107 200

164 820

165 490

107 200

1/8

1/8

1/8

1/8

TEST7.INP

TEST7.COD

TEST7.DEC

TEST7.OUT

363

116 160

178 596

179 322

116 160

1/8

1/8

1/8

1/8

TEST8.INP

TEST8.COD

TEST8.DEC

TEST8.OUT

340

108 800

167 280

167 960

108 800

2/8

2/8

2/8

2/8

TEST9.INP

TEST9.COD

TEST9.DEC

TEST9.OUT

407

130 240

200 244

201 058

130 240

2/8

2/8

2/8

2/8

TEST10.INP

TEST10.COD

TEST10.DEC

TEST10.OUT

383

122 560

188 436

189 202

122 560

2/8

2/8

2/8

2/8

TEST11.INP

TEST11.COD

TEST11.DEC

TEST11.OUT

367

117 440

180 564

181 298

117 440

2/8

2/8

2/8

2/8

TEST12.INP

TEST12.COD

TEST12.DEC

TEST12.OUT

298

95 360

146 616

147 212

95 360

2/8

2/8

2/8

2/8

TEST13.INP

TEST13.COD

TEST13.DEC

TEST13.OUT

338

108 160

166 296

166 972

108 160

2/8

2/8

2/8

2/8

TEST14.INP

TEST14.COD

TEST14.DEC

TEST14.OUT

318

101 760

156 456

157 092

101 760

(continued)

Table 5 (concluded): Location and size of speech codec test sequences

Disk No.

File Name

No. of frames

Size (bytes)

2/8

2/8

2/8

2/8

TEST15.INP

TEST15.COD

TEST15.DEC

TEST15.OUT

328

104 960

161 376

162 032

104 960

2/8

2/8

2/8

2/8

TEST16.INP

TEST16.COD

TEST16.DEC

TEST16.OUT

354

113 280

174 168

174 876

113 280

3/8

3/8

3/8

3/8

TEST17.INP

TEST17.COD

TEST17.DEC

TEST17.OUT

316

101 120

155 472

156 104

101 120

3/8

3/8

3/8

3/8

TEST18.INP

TEST18.COD

TEST18.DEC

TEST18.OUT

402

128 640

197 784

198 588

128 640

3/8

3/8

3/8

3/8

TEST19.INP

TEST19.COD

TEST19.DEC

TEST19.OUT

402

128 640

197 784

198 588

128 640

3/8

3/8

3/8

3/8

TEST20.INP

TEST20.COD

TEST20.DEC

TEST20.OUT

631

201 920

310 452

311 714

201 920

3/8

3/8

TEST21.INP

TEST21.DEC

1

320

494