## A.1 Codec parameter description

06.203GPPHalf rate speech transcodingTS

The following is a list of all the parameters which are coded for each 20 ms speech frame. The basic data rate of the speech coder is 5,6 kbps. Therefore each 20 ms speech frame consists of 112 bits. These bits are given in table A.1.

Table A.1: Codec parameter description

Parameter | No. of bits | Description | |

Frame bits: | |||

MODE | 2 | voicing mode | |

R0 | 5 | frame energy | |

LPC1 | 11 | reflection coefficient vector r1-r3 | |

LPC2 | 9 | reflection coefficient vector r4-r6 | |

LPC3 | 8 | reflection coefficient vector r7-r10 | |

INT_LPC | 1 | the soft interpolation bit for the frame | |

Subframe bits (MODE = 1,2 or 3): | |||

LAG_1 | 8 | lag for first subframe | |

LAG_2 | 4 | lag delta code for second subframe | |

LAG_3 | 4 | lag delta code for third subframe | |

LAG_4 | 4 | lag delta code for fourth subframe | |

CODE_1 | 9 | codebook, I, for first subframe | |

CODE_2 | 9 | codebook, I, for second subframe | |

CODE_3 | 9 | codebook, I, for third subframe | |

CODE_4 | 9 | codebook, I, for fourth subframe | |

GSP0_1 | 5 | {P0,GS} code for first subframe | |

GSP0_2 | 5 | {P0,GS} code for second subframe | |

GSP0_3 | 5 | {P0,GS} code for third subframe | |

GSP0_4 | 5 | {P0,GS} code for fourth subframe | |

Subframe bits (MODE=0): | |||

CODE1_1 | 7 | codebook code, I, for first subframe | |

CODE2_1 | 7 | codebook code, H, for first subframe | |

CODE1_2 | 7 | codebook code, I, for second subframe | |

CODE2_2 | 7 | codebook code, H, for second subframe | |

CODE1_3 | 7 | codebook code, I, for third subframe | |

CODE2_3 | 7 | codebook code, H for third subframe | |

CODE1_4 | 7 | codebook code, I, for fourth subframe | |

CODE2_4 | 7 | codebook code, H, for fourth subframe | |

GSP0_1 | 5 | {P0,GS} code for first subframe | |

GSP0_2 | 5 | {P0,GS} code for second subframe | |

GSP0_3 | 5 | {P0,GS} code for third subframe | |

GSP0_4 | 5 | {P0,GS} code for fourth subframe |

## A.1.1 MODE

The speech coder is defined by 4 voicing modes. MODE is a two bit code which specifies which of the four voicing modes is used at the current frame. The MODE indicates which definition of the frame bits to apply to the current frame.

## A.1.2 R0

R0 is a code which represents the average signal power of the input speech for the frame. The average signal power is computed using an analysis window which is centered over the last 100 samples of the frame.

## A.1.3 LPC1 – LPC3

The 10 reflection coefficients are vector quantized in three vector segments. The first vector segment codes reflection coefficients r1 – r3, the second vector segment codes coefficients r4-r6, the third vector segment codes coefficients r7 – r10.

## A.1.4 LAG_1 – LAG_4

LAG_1, the lag for the first subframe, can take on the value in the range of 21 to 142. Eight bits are used to encode the lag which may be fractional in value. Each of the remaining lag values (LAG_2 through LAG_4) is delta coded relative to the preceding subframe’s coded value of the lag, with a deviation of ‑8 to +7 allowable lag value levels specified by a four bit code.

## A.1.5 CODEx_1 – CODEx_4

If MODE ¹ 0, the code value for the VSELP codebook is the codeword I as derived by the codebook search procedure. If MODE=0, two VSELP codebooks are sequentially searched, with codeword I, specifying the codevector from the first VSELP codebook, assigned onto CODE1_x, and codeword H, specifying the codeword selected from the second VSELP codebook, assigned onto CODE2_x, where x is the subframe number.

## A.1.6 GSP0_1 – GSP0_4

The {P0,GS} codebook contains the values needed to determine the gain factors for the excitation vectors of a given subframe. The index of the corresponding codebook entry is assigned to GSP0_x.

The speech coder is a multimode speech coder, defined by four voicing modes:

MODE = 0 | unvoiced |

MODE = 1 | slightly voiced |

MODE = 2 | moderately voiced |

MODE = 3 | strongly voiced |

If MODE=0, the adaptive codebook (long-term predictor) and the VSELP codebook are replaced by two other VSELP codebooks.