Foreword................................................................................................................................................ 4

1....... Scope........................................................................................................................................... 5

2....... References.................................................................................................................................... 5

3....... Definitions, symbols and abbreviations........................................................................................... 5

3.1......... Definitions............................................................................................................................................................................ 5

3.2......... Symbols................................................................................................................................................................................ 6

3.3......... Abbreviations....................................................................................................................................................................... 6

4....... General......................................................................................................................................... 7

5....... Functions on the transmit (TX) side................................................................................................ 7

5.1......... Background acoustic noise evaluation............................................................................................................................ 8

5.2......... Modification of the speech encoding algorithm during SID frame generation...................................................... 11

5.3......... SID-frame encoding......................................................................................................................................................... 11

6....... Functions on the receive (RX) side............................................................................................... 12

6.1......... Averaging and decoding of the LP and fixed codebook gain parameters............................................................... 12

6.2......... Comfort noise generation and updating........................................................................................................................ 14

7....... Computational details.................................................................................................................. 15

Annex A (informative):....... Change history....................................................................................... 16

 

This Technical Specification has been produced by the 3^rd Generation Partnership Project (3GPP).

The present document defines operation of the background acoustic noise evaluation, noise parameter encoding/decoding and comfort noise generation in Mobile Stations (MSs) and Base Station Systems (BSSs) during Discontinuous Transmission (DTX) on Enhanced Full Rate speech traffic channels within the digital cellular telecommunications system.

The contents of the present document are subject to continuing work within the TSG and may change following formal TSG approval. Should the TSG modify the contents of the present document, it will be re-released by the TSG with an identifying change of release date and an increase in version number as follows:

Version x.y.z

where:

x    the first digit:

1    presented to TSG for information;

2    presented to TSG for approval;

3    or greater indicates TSG approved document under change control.

y    the second digit is incremented for all changes of substance, i.e. technical enhancements, corrections, updates, etc.

z    the third digit is incremented when editorial only changes have been incorporated in the document.

The present document gives the detailed requirements for the correct operation of the background acoustic noise evaluation, noise parameter encoding/decoding and comfort noise generation in Mobile Stations (MSs) and Base Station Systems (BSSs) during Discontinuous Transmission (DTX) on Enhanced Full Rate speech traffic channels.

The requirements described in the present document are mandatory for implementation in all GSM MSs capable of supporting the Enhanced Full Rate speech traffic channel.

The receiver requirements are mandatory for implementation in all GSM BSSs capable of supporting the Enhanced Full Rate speech traffic channel, the transmitter requirements only for those where downlink DTX will be used.

In case of discrepancy between the requirements described in the present document and the fixed point computational description of these requirements contained in GSM 06.53 [2], the description in GSM 06.53 [2] will prevail.

The following documents contain provisions which, through reference in this text, constitute provisions of the present document.

·       References are either specific (identified by date of publication, edition number, version number, etc.) or non‑specific.

·       For a specific reference, subsequent revisions do not apply.

·         For a non-specific reference, the latest version applies.  In the case of a reference to a 3GPP document (including a GSM document), a non-specific reference implicitly refers to the latest version of that document in the same Release as the present document.

[1]                          GSM 01.04: "Digital cellular telecommunications system (Phase 2+); Abbreviations and acronyms".

[2]                          GSM 06.53: "Digital cellular telecommunications system (Phase 2+); ANSI-C code for the GSM Enhanced Full Rate (EFR) speech codec".

[3]                          GSM 06.60: "Digital cellular telecommunications system (Phase 2+); Enhanced Full Rate (EFR) speech transcoding".

[4]                          GSM 06.61: "Digital cellular telecommunications system (Phase 2+); Substitution and muting of lost frame for Enhanced Full Rate (EFR) speech traffic channels".

[5]                          GSM 06.81: "Digital cellular telecommunications system (Phase 2+); Discontinuous transmission (DTX) for Enhanced Full Rate (EFR) speech traffic channels".

3.1        Definitions

For the purposes of the present document, the following terms and definitions apply:

frame: time interval of 20 ms corresponding to the time segmentation of the Enhanced Full Rate speech transcoder, also used as a short term traffic frame.

SID frame: frame characterized by the SID (SIlence Descriptor) codeword. It conveys information on the acoustic background noise.

SID codeword: fixed bit pattern for labelling a traffic frame as a SID frame.

SID field: bit positions of the SID codeword within a SID frame.

speech frame: traffic frame that cannot be classified as a SID frame.

VAD flag: Voice Activity Detection flag.

SP flag: SPeech flag.

Other definitions of terms used in the present document can be found in GSM 06.60 [3] and GSM 06.81 [5]. The overall operation of DTX is described in GSM 06.81 [5].

3.2        Symbols

For the purposes of the present document, the following symbols apply. Boldface symbols are used for vector variables.

    Unquantized LSF vector

    Quantized LSF vector

                                   mth unquantized LSF vector of the frame

                                   mth quantized LSF vector of the frame

                                    Reference LSF parameter vector

                                  Averaged LSF parameter vector

                                      Unquantized fixed codebook gain

                                      Quantized fixed codebook gain

                                   Reference fixed codebook gain

                                 Averaged fixed codebook gain

                                     Linear prediction residual signal

                                       Computed LSF parameter prediction residual

                                       Quantized LSF parameter prediction residual

                                        Computed fixed codebook gain correction factor

                                        Quantized fixed codebook gain correction factor

            

3.3        Abbreviations

For the purposes of the present document, the following abbreviations apply:

BSS                        Base Station Subsystem

DTX                       Discontinuous Transmission

LP                          Linear Prediction

LSF                        Line Spectral Frequency

LSP                        Line Spectral Pair

MS                         Mobile Station

RX                         Receive

SID                        SIlence Descriptor

TX                          Transmit

VAD                      Voice Activity Detector

For abbreviations not given in this clause, see GSM 01.04 [1].

A basic problem when using DTX is that the background acoustic noise, which is transmitted together with the speech, would disappear when the radio transmission is cut, resulting in discontinuities of the background noise. Since the DTX switching can take place rapidly, it has been found that this effect can be very annoying for the listener - especially in a car environment with high background noise levels. In bad cases, the speech may be hardly intelligible.

The present document specifies the way to overcome this problem by generating on the receive (RX) side synthetic noise similar to the transmit (TX) side background noise. The comfort noise parameters are estimated on the TX side and transmitted to the RX side before the radio transmission is switched off and at a regular low rate afterwards. This allows the comfort noise to adapt to the changes of the noise on the TX side.

The comfort noise evaluation algorithm uses the following parameters of the GSM Enhanced Full Rate speech encoder, defined in GSM 06.60 [3]:

-     the unquantized and quantized Linear Prediction (LP) parameters, using the Line Spectral Pair (LSP) representation, where the unquantized Line Spectral Frequency (LSF) vector is given by , the quantized LSF vector is given by , and the two sets of unquantized and quantized LSF vectors (one for each half of a frame) are given by , ,  and , respectively;

-     the quantized fixed-codebook gain .

The algorithm also computes the following parameters to assist in comfort noise generation:

-     the reference LSF parameter vector  (average of the quantized LSF parameters of the hangover period);

-     the averaged LSF parameter vector  (average of the LSF parameters of the eight most recent frames);

-     the reference fixed codebook gain  (average of the quantized fixed codebook gain values of the hangover period);

-     the averaged fixed codebook gain  (average of the fixed codebook gain values of the eight most recent frames);

-     the unquantized fixed codebook gain .

These parameters give information on the level (, , , ) and the spectrum (, , , , , ) of the background noise.

Two of the evaluated comfort noise parameters ( and ) are encoded into a special frame, called a Silence Descriptor (SID) frame, for transmission to the RX side. Since the reference LSF parameter vector  and the reference fixed codebook gain  can be evaluated in the same way in the encoder and decoder, as given in clause 5.1, no transmission of these parameters is necessary.

The averaged LSF parameter and fixed codebook gain values,  and , are computed in the encoder using both quantized and unquantized parameter values if the period of the eight most recent frames (the SID averaging period) is overlapping with the hangover period (the parameters from the frames overlapping with the hangover period have quantized values, while the parameters of the more recent frames of the SID averaging period have unquantized values). If the period of the eight most recent frames is non-overlapping with the hangover period, the averaged LSF parameter and fixed codebook gain values are computed using only unquantized parameter values.

The SID frame also serves to initiate the comfort noise generation on the receive side, as a SID frame is always sent at the end of a speech burst, i.e., before the radio transmission is terminated.

The scheduling of SID or speech frames on the radio path is described in GSM 06.81 [5].

5.1        Background acoustic noise evaluation

The comfort noise parameters to be encoded into a SID frame are calculated over N = 8 consecutive frames marked with VAD = 0, as follows:

The averaged LSF parameter vector  of the frame i shall be computed according to the equation:

                                                                                                                          (1)

where:

                 is the mth (unquantized) LSF parameter vector of the current frame i (n = 0);

         is the mth (quantized or unquantized) LSF parameter vector of one of the last frames (n = 1,..., 7);

      n          is the averaging period index (n = 0, 1,..., 7);

      m          is the LSF parameter vector index within a frame (1 or 2);

      i           is the frame index.

NOTE:      When the averaging is performed at the end of the hangover period (first SID update), all of the LSF parameter vectors  of the 7 previous frames (the hangover period) have quantized values, while the LSF parameter vectors  of the current frame i have unquantized values. In the subsequent SID updates, the LSF parameter vectors of the SID averaging period in the frames overlapping with the hangover period have quantized values, while the parameter vectors of the more recent frames of the SID averaging period have unquantized values.

The averaged LSF parameter vector  of the frame i is encoded using the same encoding tables that are also used by the GSM Enhanced Full Rate speech codec for the encoding of the non-averaged LSF parameter vectors in ordinary speech encoding mode, but the quantization algorithm is modified in order to support the quantization of comfort noise. The LSF parameter prediction residual to be quantized is obtained according to the following equation:

                                                                                                                                                        (2)

where:

      is the averaged LSF parameter vector at the current frame i

                 is the reference LSF parameter vector

               is the computed LSF parameter prediction residual at the current frame i

      i                  is the frame index;

NOTE:      This prediction residual is used for both halves of the frame in the quantization algorithm. The computation of the reference LSF parameter vector  is made on the basis of the quantized LSF parameters,  and , by averaging the parameters over the hangover period of 7 frames, according to the following equation:

                                                                                                                                  (3)

where:

         is the mth quantized LSF parameter vector of one of the frames of the hangover period
                         (n = 1,..., 7);

n                 is the hangover period frame index (n = 1,..., 7);

m                is the LSF parameter index within a frame (1 or 2);

k                 is the frame index.

For each comfort noise insertion period, the computation of the reference LSF parameter vector  is done only once at the end of the hangover period and for the rest of the comfort noise insertion period  will be frozen. The reference LSF parameter vector  is evaluated in the decoder in the same way as in the encoder, because during the hangover period the same LSF parameter vectors  are available at the encoder and decoder. An exception to this are the cases when transmission errors are severe enough to cause the parameters to become unusable, and the frame substitution procedure is activated (see GSM 06.61 [4]). In these cases, the modified parameters obtained from the frame substitution procedure are used instead of the received parameters.

The fixed codebook gain values shall be averaged and updated in every subframe according to the equation:

                                                                                                                                                (4)

where:

            is the (unquantized) fixed codebook gain in the current subframe (n=0);

        is the (quantized or unquantized) fixed codebook gain in one of the past

                         subframes (n = 1,..., 28);

      n                 is the averaging period index (n = 0, 1,..., 28);

NOTE:      When the averaging is started at the end of the hangover period (first SID update), all of the fixed codebook gains  of the 28 (n=1,..., 28) previous subframes (the hangover period) have quantized values, while the fixed codebook gains  of the current subframe has an unquantized value. In the subsequent SID updates, the fixed codebook gain values of the SID averaging period in the subframes overlapping with the hangover period have quantized values, while the parameter vectors of the more recent subframes of the SID averaging period have unquantized values.

Since most parts of the subframe processing clause in the encoder are switched off when the SP flag = "0" (to minimize the average complexity of the speech encoder algorithm), the unquantized fixed codebook gain is not directly available for gain averaging. Due to this, the unquantized fixed codebook gain is separately computed, based on the energy of the LP residual signal in each subframe, according to the following equation:

                                                                                                                                         (5)

where:

                  is the unquantized fixed codebook gain of the current subframe j;

           is the lth sample of the LP residual in the current subframe j;

      j                        is the subframe index (j = 1,..., 4);

      l                        is the sample index (l = 1,..., 40).

NOTE:      The computed energy of the LP residual signal is divided by the value of 10 to yield the energy for one excitation pulse, since during comfort noise generation, the subframe excitation signal (pseudo noise) has 10 non-zero samples, whose amplitudes can take values of +1 or ‑1.

The averaged fixed codebook gain value  of the current subframe is encoded using the non‑averaged fixed codebook gain values in ordinary speech encoding mode, but the quantization algorithm is modified in order to support comfort noise quantization. The fixed codebook gain correction factor  to be quantized is obtained according to the following equation:

                                                                                                                                                                  (6)

where:

             is the averaged fixed codebook gain value in the current subframe;

                is the reference fixed codebook gain;

The computation of the reference fixed codebook gain  is made on the basis of the quantized fixed codebook gain parameters , by averaging the parameter values over the hangover period of 7 frames according to the following equation:

                                                                                                                               (7)

where:

      is the quantized fixed codebook gain parameter value in subframe j of one of the frames of the hangover period (n = 1,..., 7);

      n                       is the hangover period frame index (n = 1, 2,..., 7);

      k                       is the frame index;

      j                        is the subframe index (j = 1,..., 4);

For each comfort noise insertion period, the computation of the reference fixed codebook gain  is done only once at the end of the hangover period and for the rest of the comfort noise insertion period  will be frozen. The reference fixed codebook gain  can be evaluated in the decoder in the same way as in the encoder, because during the hangover period the same quantized fixed codebook gain values  are available at the encoder and decoder. An exception to this are the cases when transmission errors are severe enough to cause the parameters to become unusable, and the frame substitution procedure is activated (see GSM 06.61 [4]). In these cases, the modified parameters obtained from the frame substitution procedure are used instead of the received parameters.

The hangover period is defined in GSM 06.81 [5]. It is a period added at the end of a speech burst in which no voice activity is detected (VAD flag = "0"), but the speech encoder stays for the processing of 7 speech frames in speech encoding mode (SP flag = "1"). This hangover period and the first SID frame are used for averaging the comfort noise parameters contained in the first SID frame.

5.2        Modification of the speech encoding algorithm during SID frame generation

When the SP flag is equal to "0" the speech encoding algorithm is modified in the following way:

-     The non-averaged LP parameters which are used to derive the filter coefficients of the filters  and  of the speech encoder are not quantized.

-     The open loop pitch lag search is performed, but the closed loop pitch lag search is inactivated. The adaptive codebook gain is set to zero.

-     No fixed codebook search is made. In each subframe the pulse positions and signs of the fixed codebook excitation are locally generated using uniformly distributed pseudo random numbers. The excitation pulses take values of +1 and ‑1 when comfort noise is generated. The fixed codebook comfort noise excitation generation algorithm is defined in clause 6.2.

-     The memory of weighting filter  is set to zero, i.e., the memory of  is not updated.

-     The ordinary LP parameter quantization algorithm is inactive. At the end of the hangover period the reference LSF parameter vector  is calculated as defined in clause 5.1. For the rest of the comfort noise insertion period  is frozen. The averaged LSF parameter vector  is calculated each time a new SID frame is to be sent to the Radio Subsystem. This parameter vector is encoded into the SID frame as defined in clause 5.1.

-     The ordinary fixed codebook gain quantization algorithm is inactive. At the end of the hangover period the reference fixed codebook gain  is calculated as defined in clause 5.1. For the rest of the comfort noise insertion period  is frozen. The averaged fixed codebook gain value  is calculated each time a new SID frame is to be sent to the Radio Subsystem. This gain value is encoded into the SID frame as defined in clause 5.1.

-     The predictor memories of the ordinary LP parameter quantization and fixed codebook gain quantization algorithms are reset when SP flag = "0", so that the quantizers start from their initial states when the speech activity begins again.

-     The computation of the unquantized fixed codebook gain is performed based on the energy of the LP residual signal.

5.3        SID-frame encoding

The SID-frame encoding algorithm exploits the fact that only some of the 244 bits in a frame are needed to code the comfort noise parameters. The other bits can then be used to mark the SID-frame by means of a fixed bit pattern, called the SID code word.

The quantization indices of the LP parameters are replaced by the quantization indices derived from the averaged LSF parameter vector . The encoding of the quantization indices is defined in clause 5.1.

The fixed codebook gain quantization indices are replaced by the quantization index derived from the averaged fixed codebook gain value , encoded as defined in clause 5.1, repeated four times inside the frame.

The SID code word consists of 95 bits which are all set to one. The bits of the SID code word are inserted in the SID field as defined in table 1. All of the bits in the SID field are in the error protection Class I.

The remaining bits in the SID frame are set to zero. The use of these bits is for further study.

Table 1: SID codeword

 

The parameters in table 1 are defined in GSM 06.60 [3].

The situations in which comfort noise shall be generated on the receive side are defined in GSM 06.81 [5]. Generally speaking, the comfort noise generation is started or updated whenever a valid SID frame is received.

6.1        Averaging and decoding of the LP and fixed codebook gain parameters

When speech frames are received by the decoder the LP and the fixed codebook gain parameters of the last seven speech frames shall be kept in memory. The decoder counts the number of frames elapsed since the last SID frame was updated and passed to the RSS by the encoder. Based on this count, the decoder determines whether or not there is a hangover period at the end of the speech burst (if at least 31 frames have elapsed since the last SID update when the first SID frame after a speech burst arrives, the hangover period has existed at the end of the speech burst).

As soon as a SID frame is received, and the hangover period is detected at the end of the speech burst, the stored LP and fixed codebook gain parameters shall be averaged to obtain the reference LSF parameter vector  and the reference fixed gain codebook value . The reference LSF parameter vector and the reference fixed codebook gain value shall be frozen and used for the actual comfort noise insertion period.

The averaging procedure for obtaining the reference parameters is as follows:

-     when a speech frame is received, the LSF and fixed codebook gain parameters are decoded and stored in memory;

-     when the first SID frame is received, and the hangover period is detected at the end of the speech burst, the stored LSF and fixed codebook gain parameters are averaged in the same way as in the speech encoder as follows (see also clause 5.1):

                                                                                                                                  (8)

where:

         is the mth quantized LSF parameter vector of one of the frames of the hangover period
                         (n = 1,..., 7);

      n                       is the hangover period frame index (n = 1, 2,..., 7);

      m                      is the LSF parameter index within a frame (1 or 2);

      k                       is the frame index.

      and

                                                                                                                               (9)

where:

      is the quantized fixed codebook gain parameter value in subframe j of one of the                                                          frames of the hangover period (n = 1,..., 7);

      n                       is the hangover period frame index (n = 1,..., 7);

      k                       is the frame index;

      j                        is the subframe index (j = 1,..., 4).

Once the reference LSF parameter vector has been computed, the averaged LSF parameter vector  of the frame i (encoded into the SID frame) can be reproduced at the decoder each time a SID update frame is received, according to the equation:

                                                                                                                                                       (10)

where:

      is the quantized, averaged LSF parameter vector at the current frame i to be used for comfort noise                          generation;

                 is the reference LSF parameter vector;

               is the received quantized LSF parameter prediction residual at the current frame i;

      i                  is the frame index.

The averaged fixed codebook gain  of the frame i (encoded into the SID frame) can be similarly reproduced at the decoder each time a SID update frame is received, according to the following equation:

                                                                                                                                                     (11)

where:

      is the averaged fixed codebook gain value in the current frame i to be used for comfort noise                                     generation;

                is the reference fixed codebook gain;

             is the received quantized fixed codebook gain correction factor in the current frame i;

      i                  is the frame index.

6.2        Comfort noise generation and updating

The comfort noise generation procedure uses the GSM Enhanced Full Rate speech decoder algorithm defined in GSM 06.60 [3].

When comfort noise is to be generated, the various encoded parameters are set as follows:

In each subframe, the pulse positions and signs of the fixed codebook excitation are locally generated using uniformly distributed pseudo random numbers. The excitation pulses take values of +1 and ‑1 when comfort noise is generated. The fixed codebook comfort noise excitation generation algorithm works as follows:

      for (i = 0; i < 40; i++)      code[i] = 0;

      for (i = 0; i < 10; i++) {

      j = random(4);

      idx = j * 10 + I;

      if (random(2) == 1)   code[idx] = 1;

      else                               code[idx] = ‑1;

      }

where:

      code[0..39]  fixed codebook excitation buffer;

      random(4)       generates a random integer value, uniformly distributed between 0 and 3;

      random(2)       generates a random integer value, uniformly distributed between 0 and 1.

The fixed codebook gain values are those received in the SID frame.

The adaptive codebook gain values in each subframe are set to 0.

The pitch delay values in each subframe are set to 40.

The LP filter parameters used are those received in the SID frame.

The predictor memories of the ordinary LP parameter and fixed codebook gain quantization algorithms are reset when SP flag = "0", so that the quantizers start from their initial states when the speech activity begins again.

With these parameters, the speech decoder now performs the standard operations described in GSM 06.60 [3] and synthesizes comfort noise.

Updating of the comfort noise parameters (fixed codebook gain and LP filter parameters) occurs each time a valid SID frame is received, as described in GSM 06.81 [5].

When updating the comfort noise, the parameters above should preferably be interpolated over the SID update period to obtain smooth transitions.

A low level description has been prepared in form of an ANSI C source code which is part of GSM 06.53 [2].

 

 

Change history
SMG No.	TDoc. No.	CR. No.	Clause affected	New version	Subject/Comments
SMG#22				4.0.1	ETSI Publication
SMG#20				5.1.2	Release 1996 version
SMG#27				6.0.0	Release 1997 version
SMG#29				7.0.0	Release 1998 version
				7.0.1	Version update to 7.0.1 for Publication
SMG#31				8.0.0	Release 1999 version
				8.0.1	Update to Version 8.0.1 for Publication

 

 

Change history
Date	TSG #	TSG Doc.	CR	Rev	Subject/Comment	Old	New
03-2001	11				Version for Release 4		4.0.0
06-2002	16				Version for Release 5	4.0.0	5.0.0
12-2004	26				Version for Release 6	5.0.0	6.0.0
06-2007	36				Version for Release 7	6.0.0	7.0.0
12-2008	42				Version for Release 8	7.0.0	8.0.0