KR20050027272A - Speech communication unit and method for error mitigation of speech frames - Google Patents

Abstract

A speech communication unit (100) comprising a speech encoder (134) capable of representing an input speech signal, the speech encoder (134) comprising a transmission path (281) for transmitting a number of speech frames to a speech decoder, the speech encoder (134) characterised by a virtual transmission path (282) for transmitting one or more references for a number of speech frames transmitted in the transmission path (281) wherein the one or more references relate to an alternative speech frame within the number of speech frames transmitted on the transmission path (281) to be used as a replacement frame when a frame is received in error. The speech communication unit provides at least the advantage that a more accurate replacement frame mechanism is provided, thereby reducing the risk of undesirable artefacts being audible in recovered speech frame.

Description

Speech communication unit and method for error mitigation of speech frames

The present invention relates to speech coding and methods for improving the performance of speech codecs in speech communication units. The present invention can be used for error mitigation in speech codecs, but is not limited thereto.

Many current voice communication systems, such as the Global System for Mobile Communications Globalization System (GSM) cellular telephony for personal mobile wireless users and the Global System for Terrestrial Trunked RAdio (TETRA) systems, are used to encode and decode speech patterns. Use speech-processing units. In such communication systems, the speech encoder of the transmission unit converts the analog speech pattern into a digital format suitable for transmission. The speech decoder of the receiving unit converts the received digital speech signal into an audible analog speech pattern.

Since the frequency spectrum for the wireless voice communication systems is an important resource, it is desirable to limit the channel bandwidth used by the speech signals in order to maximize the number of users per frequency band. Thus, a first purpose of using speech coding techniques is to reduce the occupied capacity of speech patterns as much as possible by the use of compression techniques without loss of fidelity.

In the context of voice and data communication systems, another approach is to provide substantially less protection on speech signals when compared to similar data signals. This approach causes relatively more errors in the speech packet than the data packets and increases the risk of corruption of the overall speech packets.

In speech decoders, it is common to use error mitigation techniques to improve the performance of the speech communication unit in the following event.

The event may (i) provide too many bit errors in the received speech frame;

(Ii) Data packets (which may include speech information) in the network-based Internet Protocol (IP) are lost.

"Bad-frame" mitigation techniques are necessary to minimize the audible effect of frames received in an error, where "error reception" is taken here to mean either received with errors or not received at all. These techniques reproduce an estimate of the lost speech frame rather than injecting either silence or noise into the decoded speech. The techniques typically include using the statistical static properties of speech. A single frame of error is generally adequately evaluated by replacing the frame with similar parameters including energy, pitch, spectrum and speech from the previous frame. However, speech is truly unchanged, for example speech starts and burst sounds are very short events. Thus, this simple "alternate" technique sometimes causes abnormalities, thus forming undesirable workpieces.

In an ideal world, it would be desirable to interpolate data from either aspect of transmission interruption. That is, it is desirable not only to take data after a bad-frame sequence, but also to take previous data and interpolate between them. However, the method is not allowed in voice communication systems because it introduces an undesirable delay.

If several bad-frames are received, then the energy of the speech signals is often reduced to zero after several frames. Often the "voice" parameter is included because it is available for repeated changes depending on whether speech is spoken or not. In principle, pronounced speech is preferred for merely repeating periodic components. In contrast, for unpronounced speech, it is desirable to produce similar audio spectrum and similar energy not too periodically.

The inventors of the present invention have recognized and understood the limitations of using a simple "alternative" frame mechanism as a bad-frame mitigation method. In particular, the inventors are recognized to replace a frame with a truly suitable frame only in rare cases. Also, if multiple frames receive an error that can occur primarily on poor quality wireless communication links, the alternative frame mechanism is less permissible.

1 is a block diagram of a wireless communication unit including a speech encoder provided to support various concepts of a preferred embodiment of the present invention.

2 is a block diagram of a code exited linear predictive speech corder adapted to support various inventive concepts of a preferred embodiment of the present invention.

3 illustrates the use of a reference mechanism indicated by an alternative virtual transmission path such that replacement frames are selected from a number of other frames, in accordance with a preferred embodiment of the present invention.

4 illustrates an improved use of an alternative virtual transmission path to handle multiple errors occurring in the primary transmission path, in accordance with a preferred embodiment of the present invention.

Thus, there is a need to provide an improved error mitigation technique when using the speech codecs to eliminate at least some of the disadvantages described above.

In a first aspect of the invention, a speech communication unit is provided according to claim 1.

In a second aspect of the invention, a speech communication unit is provided according to claim 11.

According to a third aspect of the invention, a method for performing bad-frame error mitigation in a voice communication unit is provided according to claim 13.

In a fourth aspect of the invention, a speech communication unit is provided according to claim 14.

In a fifth aspect of the invention, a wireless communication system is provided according to claim 15.

Other aspects of the invention are defined in the dependent claims.

In summary, it is an object of the present invention to provide a communication unit comprising a speech codec and method for performing bad-frame error mitigation that at least alleviates some of the above-mentioned disadvantages associated with current bad-frame error mitigation techniques. This is accomplished by using a reference / pointer transmitted on the virtual transmission path to send speech frames on the transmission path and to point to alternative alternative speech frames to be used by the speech decoder if a speech frame is received in error on the transmission path. do. Ideally by using different error statistics, for example an additional virtual transmission path using a separate FEC method, the reference / pointer will not be affected by the same errors as the speech frame it refers to. In addition, a buffering technique is used in the encoder to select an alternative speech frame from a plurality of previously transmitted speech frames that exhibit similar characteristics to the selected speech frame to be referenced.

Exemplary embodiments of the invention will be described immediately with reference to the accompanying drawings.

Referring now to FIG. 1, shown is a block diagram of a wireless subscriber unit, hereinafter referred to as a mobile station (MS) 100, adapted to support the inventive concepts of preferred embodiments of the present invention. The MS 100 has an antenna 102 that is preferably coupled to a duplex filter, antenna switch, or circulator that provides isolation between the receiver and transmitter chains within the MS 100. .

As is known in the art, the receiver chain typically includes a scanning receiver front-end circuit 106 (which effectively provides reception, filtering and intermediate or base-band frequency conversion). The scanning front-end circuit is connected in series to a single processing function 108. The output from the signal processing function is provided via a speech processing unit 130 to a suitable output device 110, such as a speaker.

The speech processing unit 130 includes a speech encoding function 134 to encode the speech of the user in a format suitable for transmission over the transmission medium. Speech-processing unit 130 includes speech decoding function 132 to decode the received speech in a format suitable for output via output device (speaker) 110. Speech-processing unit 130 is operatively coupled to memory unit 116, and timer 118 via controller 114. In particular, the operation of speech-processing unit 130 has been provided to support the inventive concepts of the preferred embodiment of the present invention. In particular, speech-processing unit 130 has been provided for selecting an alternate speech frame from a plurality of previously transmitted speech frames. Speech processing unit 130, or signal processor 108, then refers to transmitting the reference / pointer signal (indicative of the selected alternate speech frame) of the alternative virtual transmission path to the primary transmission path. Adaptation of the speech-processing unit 130 is further described with reference to FIG. 2.

To complete, the receiver chain includes a received signal strength indicator (RSSI) circuit 112 (although the RSSI circuit 112 is shown defective at the scanning receiver front end 106 if placed somewhere in the receiver chain). The RSSI circuit is coupled to the controller 114 for maintaining overall subscriber unit control. The controller 114 is also coupled to the scanning receiver front-end circuit 106 and the signal processing function 108 (generally realized by the DSP). Thus, the controller 114 may receive a bit error rate (BER) or frame error rate (FER) from the recovered information. The controller 114 is coupled to the memory device 116 to store operating situations such as decryption / encoding functions. The timer 118 is generally coupled to the controller 114 to control the operating timings (transmission or reception of time-dependent signals) within the MS 100. In the context of the present invention, timer 118 indicates the timing of speech signals in the transmit (encoded) path and / or receive (decoded) path.

In the context of a transmission chain, this essentially includes an input device 120 such as a microphone converter coupled in series via a speech encoder 134 to a transmitter / modulation circuit 122. Thereafter, any transmit signal to be radiated from antenna 102 is passed through power amplifier 124. The transmit / modulation circuit 122 and the power amplifier 124 are operatively responsive to the controller, and the output from the power amplifier is coupled to the duplex filter or transmitter 104. Transmitter / modulation circuitry 122 and scanning receiver front-end circuitry 106 include frequency up-conversion and frequency down-conversion functions (not shown).

Of course, the various components within the MS 100 may be arranged in any suitable functional topology to enable use of the concepts of the present invention. In addition, the various components in the MS 100 are realized in discrete or integrated component form, and the ultimate structure is simply any choice.

It is within the consideration of the present invention that the preferred buffering or processing of speech signals can be realized in software, firmware or hardware using a software processor (or digital signal processor (DSP)) that performs the speech processing function.

Referring to FIG. 2, a block diagram of a sign excited linear prediction (CELP) speech encoder 134 is shown in accordance with a preferred embodiment of the present invention. The acoustic input signal to be analyzed is applied to the speech encoder 134 in the microphone 202. The input signal is applied to the filter 204. The filter 204 will generally exhibit band pass filter characteristics. However, if speech bandwidth is already adequate, filter 204 may include a direct wired connection.

The analog speech signal from the filter 204 is converted into a sequence of N pulse samples, and the magnitude of each pulse sample is then known in the art, then the digital of the analog to digital (A / D) converter 208. It is represented by a sign. The sample clock SC is generated together with the frame clock FC.

The digital output of the A / D 208, which can be represented as the input speech vector s (n), is applied to the coefficient analyzer 210. This input speech vector s (n) is obtained repeatedly in frames, i.e. blocks of time, length determined by the frame clock FC, as is known in the art.

For each block of speech, a set of linear prediction coding (LPC) parameters are formed by coefficient analyzer 210 in accordance with a preferred embodiment of the present invention. The generated speech encoder parameters include LPC parameters, long term predictor (LTP) parameters, excitation gain factor G ₂ (along with the best stochastic code excitation codeword I). The speech coding parameters are applied to the multiplexer 250 and transmitted over the channel for use by speech synthesis in the decoder. Input speech n (s) is also applied to subtractor 230, the function of which is described later.

In the conventional CELP coder of FIG. 2, the codebook search controller 240 selects the best indices and takes input speech samples from the probabilistic codebook in block 214 and the adaptive codebook in block 216. Form the maximum weighting error of the summed select excitation vector used to represent. The outputs of the probabilistic code list 214 and the adaptive code list 216 are input to the gain functions 222 and 218, respectively. Gain adjustment outputs are summed in summer 220 and input to LPC filter 224 as is known in the art.

First, the adaptive code book or long term predictor component is 1 (n) calculated. This is characterized by the delay and gain factor 'G ₁ '.

For each stochastic code excitation vector u _i (n), a reconstructed speech vector s' _i (n) is generated for comparison to the input speech vector s (n). Gain block 222 scales the excitation gain factor 'G ₂ ' and summing block 220 adds to the adaptive code block component. The gain can be optimized with a search for the best excitation codeword I, precomputed by coefficient analyzer 210 and used to analyze all excitation vectors or generated by codelock search controller 240.

The scale excitation signal G ₁ 1 (n) + G ₂ u _i (n) is generated by a linear prediction coding filter 224 constituting a short-term predictor (STP) filter to produce a reconstructed speech vector s' _i (n). Is filtered. The reconstructed speech vector s' _i (n) for the ith excitation vector is compared with the same vector of the input speech vector s (n) by subtracting these two signals in subtractor 230.

The difference vector e _i (n) represents the difference between the blocks of speech and the reconstructed blocks. The difference vector is perceptually weighted by the weighting filter 232 using the weighting filter parameters WTP generated by the coefficient analyzer 210. Predictive weighting highlights frequencies and other frequencies where errors are very perceptually important to the human ear.

The energy calculator function in the code list search controller 240 calculates the energy of the weighting difference vector e ' _i (n). The codebook search controller compares the i th error signal for the excitation vector u ' _i (n) provided for the previous error signals to determine the excitation vector that produces the minimum error. The sign of the i th excitation vector with the least error is output through the channel as the best excitation sign I.

A copy of the scaled excitation G _i 1 (n) + G ₂ u _I (n) is stored in the long term predictor memory of 216 for future use.

Alternatively, codelock search controller 240 may determine a particular codeword that provides an error signal with some predetermined reference, such as meeting a predetermined error threshold.

A more detailed description of a typical speech coding unit function can be found in "Digital Speech Coding for Low Bit Rate Systems" by A.M.Kondoz, published by John Wiley in 1994.

In a preferred embodiment of the present invention, the error mitigation technique is applied to speech frames following multiplexer 250. The present invention utilizes another, preferably parallel virtual transmission path 282 used to transmit one pointer to a previously encoded speech frame transmitted from an encoder on the main transmission path 281.

In the context of the present invention, the expression 'virtual' is defined as a transmission path provided from the encoder to the decoder in addition to the main transmission path supporting speech communication.The 'virtual' transmission path is the same time frame or in the same time frame or time division multiplexing method. It can be placed in multiple frames or through another communication route, eg a VoIP system, ideally by using an additional transmission path using other error statistics, for example a separate FEC method, the reference / pointer You will not be affected by the same errors as the speech frame you reference.

One notable difference to known encoding devices is that there is a second minimization section following the multiplexing operation. The circuit evaluates the speech parameter data held in the buffer and selects the one closest to the current speech frame.

In one improved embodiment, the parallel virtual transmission path uses forward error correction (FEC) protection that is different from that used for the main transmission path by the speech encoder. In this way, by using independent FEC paths, speech data packets suffer from other error statistics. This difference between the main and parallel virtual transmission paths improves the robustness to errors.

The multiplexer 250 outputs data packets / frames to a buffer 260 that holds previously multiplexed frames. Demultiplexer 270 evaluates buffer frames of the multiplexed signal held in buffer 260. In this regard, demultiplexer 270 separates excitation parameters 274 from LPC parameters 272. Note that the memory of the long term predictor used to generate the excitation parameters should be the same as the long term predictor 216 at the start of the frame.

For each block of multiplexed speech, therefore, a set of linear predictive coding (LPC) parameters are generated for current frames and previous frames. In a preferred embodiment of the present invention, each quantized LPC parameters and excitation parameters form reconstructed speech vectors s' _j (n) for the j th previous frame of buffered data. These are compared to previously buffered speech vectors by subtracting these two signals in subtractor 262.

The difference vector e _j (n) represents the difference between the original buffer blocks of the speech and the previously buffered blocks. The difference vector is weighted percent by LPC weighting filter 264. As indicated, perceptual weighting attenuates the frequency, other frequencies, with perceptually more significant errors in the human ear.

An energy calculator function inside the sign search controller 266 calculates the energy of the weighting difference vector e ' _j (n). The code list search controller 266 compares the j th error signal for the excitation vector u _j (n) provided for the previous error signals to determine the excitation vector that produces the minimum error. The code list search controller 266 selects an 'index to the best frame data' to provide the minimum weighted error. The encoder then sends a 'pointer' to the decoder for the previous frame determined when providing the minimum weighted error between each speech frame of the main transmission path and itself.

Essentially, the referenced speech frame (ideally separate from the currently transmitted frame by time or frame number) is the frame in the current moving window of speech that is most closely similar to the frame that was encoded by the encoder upon perceptually weighted error discrimination. Configure Therefore, it represents the best match (pointer) to the current frame for use in the error mitigation process if the frame is received in error. This representation, or pointer, is described in more detail with respect to FIG. 3.

Referring to Figure 3, buffer timing diagram 300 illustrates a preferred process of the present invention. The timing diagram shows frame 0 310 which was determined to be an error and was received at the speech decoder. The decoder then evaluates the alternative virtual transmission path to determine the most suitable frame to replace frame 0 310. As shown in FIG. 3, an alternative virtual transmission path includes points for frame-4 320 in a preferred replacement of frame 0 310. By replacing frame 0 310 with frame -4 320, it has a minimal effect on speech quality during speech decoding.

The inventors of the present invention have recognized and used the fact that the preceding frames were all typically spoken by the same speaker (ie, speech frames would represent similar pitch and format positions). Therefore, there is a high likelihood that similar previous speech frames may be found in the current speech frame.

In accordance with a preferred embodiment of the present invention, the least perceptual error provides a set of parameters for each frame in memory, whereby a weighted segment signal-to-noise (SEGSNR) or average weighted SNR for each buffered frame Is found by evaluating. Preferably, the segment is defined at the speech codec sub-frame level.

This decision is made at the encoder. If there is a small pitch error, it is considered that very different SEGSNR values may appear. This is because the source speech and the buffered signal can be quickly shifted out of phase. Thus, in an improved embodiment of the present invention, it is proposed to search for pitch periods, ie +/- 5%, for frames buffered using subsample resolution (typically 1/3 or 1/4 samples). To get the highest SEGSNR value.

In another refinement of the invention, if the frame itself is received in error, the frame used to mitigate the bad reception of the frame itself is the best source of speech information for the current frame received with the error shown in FIG. to be. Thus, Figure 4 shows a timing diagram indicating how multiple errors are handled. Data from frame 0 410 is known as an error. The proposed error mitigation process uses an alternative virtual transmission path pointing to data frame 4 420 as a suitable replacement. However, data frame-4420 is determined to be an error. In this case, the pointer points to the data from frame -6 430 as one frame that is the most similar frame to modulated frame -4 420. Therefore, frame-6450 is used to replace frame-4420 and is suitable to replace frame-1410. In this way, multiple frame errors can be adjusted to overcome the problem of out of memory references.

This can ultimately lead to references (pointers) that effectively lead out of the storage window. However, this is not required as a problem if the error values in the window are updated by removing the need for multiple references.

Optionally, if replaced frames are stored in the buffer, before frame -4 420 is the current frame, the buffer is replaced by frame -6 430 (so frame -2) in the buffer, so that the buffer always contains only available data. do.

In summary, a reference or pointer is sent to the decoder of another bit stream for the main bit stream. The reference or pointer points to a previously transmitted frame that best matches the currently transmitted frame. The reference or pointer is preferably transmitted in a parallel bit stream. If a frame is received as an error in the speech decoder, a reference or pointer is used in the frame replacement error mitigation process. Thus, frame alleviation is improved by extending the immediately preceding or next successive frame replacement mechanism known to any frame from multiple frames. In this regard, the number of frames used in the process is limited only by the buffering / storage mechanism and / or processing power required to determine the minimum weighted error frame.

As indicated, the buffering / storing process of speech parameters of the speech encoder is performed over the number of frames. For example, for a GSM Enhanced Full Rate (EFR) codec of <12 kb / sec, the storage for 3 seconds of speech is only 5 kbytes. The most difficult task is to identify the closest frame match from one hundred fifty possible frames. Thus, in one embodiment of the present invention, the minimum weighted error selection technique described above may be applied to a subset or parameters of parameters derived from synthesized speech, rather than all of the parameters of a speech encoder frame. In other words, the energy and LPC filter parameters (LSF) of the synthesized speech frame (speech parameter derived from the synthesized speech calculated in both the encoder and the decoder) are referenced rather than precise encoder parameters in order to save memory and computation. Or pointers).

In this regard, since the speech frame contains many parameters, the proposed technique can be applied to any number of parameters. Embodiments of the parameters in the CELP encoder are as follows:

(1) line spectral pairs (LSP) representing LPC parameters;

(2) long term predictor (LTP) lag lagging for subframe-1;

(3) LTP gain for subframe-1;

(4) codelock index for subframe-1;

(5) codelock gain for subframe-1;

(6) long term predictor lag for subframe-2;

(7) LTP gain for subframe-2;

(8) codelock index for subframe-2;

(9) codelock gain for frame-2;

(10) long term predictor lag for subframe-3;

(11) LTP gain for subframe-3;

(12) codelock index for subframe-3;

(13) codelock gain for subframe-3;

(14) long term predictor lag for subframe-4;

(15) LTP gain for subframe-4;

(16) codelock index for subframe-4; or

(17) Codelock gain for subframe-4.

It is within the consideration of the present invention that a pointer can be sent by referring to the set of LSPs from previous frames to match that of the current frame rather than the entire set of parameters. Optionally, it is possible to have a pointer to each of the plurality of parameters.

In a wireless communication system, a parallel virtual transmission path involves transmitting a block coded reference word in unprotected bits of data cost (where seven bits are sufficient to support a 128 frame buffer, such as approximately 2.5 seconds). Do). It can be coded with a 15 bit BCH block code that provides up to 2 bit error correction (with an equivalent rate of 75 bits / sec).

Optionally, it is contemplated that alternative virtual transmission paths may provide a combination of error correction and error detection functions. Error detection is useful because poor reception of a reference causes poor alleviation. In the case of an incorrectly received reference word, the method may default to the previous frame repetition. The channel rate of 75 bits / sec reduces only the gross bit-rate of GSM full-rate channels of 22.8 Kbit / sec to 22.725 Kbit / sec, resulting in negligible loss of sensitivity.

In other embodiments, such as voice over an Internet Protocol (VoIP) communication link, an alternative virtual transport path may be achieved by sending multiple packet streams. In this environment, it is desirable that the total traffic does not increase substantially because this increases the probability of packet dropping.

The preferred mechanism sends a reference to previous frames as described above only if transmissions occur and speech is not fixed. When speech is fixed, and when the prior art works relatively well, references are not sent. In this way the packet network is not overloaded and most of the performance gains are achieved. The degree to which the speech signal is static can be generated as a variable that can be adjusted to improve the reproduced quality in case of lost packets.

The decoder functionality is substantially the inverse of the encoder (without additional circuitry following the multiplexer) and therefore is not described in detail here. A functional description of a typical speech decoding unit can also be found in "Digital Speech Encoding for Slow Bit Communication Systems" by A.M.Kondoz, published by John Wiley in 1994. In the decoder, the decoder follows the standard decoding process until it determines a bad-frame. When a bad-frame is detected, the decoder evaluates the alternative virtual transmission path to determine the alternative frame indicated by each reference / pointer. The decoder then searches for 'similar' frames as indicated by the reference / pointer transmission. The previously indicated frame is used to replace the received frame to synthesize speech.

Advantageously, the inventive concepts described herein can be retrofitted into conventional codecs by obtaining bits from an already configured FEC method.

It is within the scope of the present invention that any speech processing circuit is preferred from the inventive concept described above.

It will be appreciated that the bad-frame error mitigation mechanism described above provides at least the following advantages:

(1) A more accurate frame mechanism is provided, which reduces the risk of undesirable hardening that may be audible in recovered speech frames.

(2) An alternative virtual transmission path can be retrofitted to conventional codecs, for example by obtaining bits from an already configured FEC method.

(3) When references to previous frames are sent only when transitions occur and speech is not fixed, conventional bad-frame error mitigation techniques are used to mitigate any additional data required by the present invention.

(4) By cross-referencing the frames referred to in this method with the data received for a given frame, erroneously received parameters can be detected.

Although the preferred embodiment has discussed the CELP encoder in the application of the present invention, it is contemplated by the inventors that other speech processing units in which transmission errors may occur may be desirable from the inventive concepts contained herein. The concepts of the invention described herein include the Universal Interchange of Mobile Telecommunications System (UMTS) units, Global System for Mobile Communications Globalization System (GSM), Terrestrial Relay Radio (TETRA) communication units, Information and Signaling Standards. A particular use is found in speech processing units for wireless communication units such as (DIIS), voice over voice (VoIP) units, and the like.

Device of the Invention

The speech communication unit includes a speech encoder that can represent an input speech signal. The speech encoder includes a transmission path for transmitting a number of speech frames to the speech decoder. The speech encoder further includes a virtual transmission path for transmitting one or more references to the plurality of speech frames transmitted in the transmission path. One or more references relate to alternative speech frames in a number of speech frames transmitted on a transmission path to be used as replacement frames when the frame is received in error.

A speech communication unit comprising a speech communication unit, for example a speech encoder, comprises a speech decoder provided for receiving a plurality of speech frames on a virtual transmission path. One or more references relate to an alternative frame in multiple speech frames received on a transmission path to be used as a replacement frame when the frame is received in error.

Method of the invention

A method of performing bad-frame error mitigation comprises transmitting a plurality of speech frames on a transmission path to a speech decoder by a speech encoder of a switch communication unit. The speech encoder transmits one or more references to multiple speech frames transmitted on the transmission path on the virtual transmission path. One or more references relate to alternative speech frames in a number of speech frames transmitted on a transmission path to be used as replacement frames when the frame is received in error.

In this way, an improved replacement frame from multiple speech frames can be selected when the speech frame is received in error.

Thus, a bad-frame error mitigation technique, and associated speech communication units and circuits, have been described that substantially eliminate at least some of the disadvantages noted above with known error mitigation techniques.

Claims (15)

A speech communication unit comprising a speech encoder capable of representing an input speech signal, the speech encoder comprising the speech encoder 134 comprising a transmission path 281 for transmitting a plurality of speech frames to the speech decoder. In the speech communication unit 100, The speech encoder 134 features a virtual transmission path 282 for transmitting one or more references to multiple speech frames transmitted in the transmission path 281, wherein the one or more references receive a frame in error. Speech communication unit, characterized in that it relates to an alternative speech frame within the number of speech frames transmitted on said transmission path (281) to be used as a replacement frame. The speech encoder 134 of claim 1, A multiplexer (250) for multiplexing the plurality of speech frames; A buffer 260 operatively coupled to the multiplexer 250 for storing multiplexed speech data; And A processor (130, 170) operatively coupled to the buffer (260) for characterizing a current speech frame of the buffer 260 and for selecting an alternative speech frame that exhibits features similar to the speech frame, wherein A speech communication unit, characterized in that said processor (130, 270) is sent to a decoder of a virtual transmission path (282). 3. The LPC parameters of the buffered speech frame as recited in claim 2, wherein the processor includes a demultiplexer function 270 for accessing one or more speech frames in the buffer 260, and for selecting a speech frame exhibiting similar characteristics. Speech communication unit, characterized in that separating excitation parameters 274 from 272. Speech communication unit according to any of the preceding claims, characterized in that the virtual transmission path (282) is included in the same bit stream of the transmission path (281). 5. The transmission path 281 according to any one of the preceding claims, wherein the transmission path 281 uses a first forward error correction protection method and the virtual transmission path 282 is different from that used in the transmission path 281. A speech communication unit, characterized in that using a second forward error correction protection method. 6. Speech communication unit according to claim 2 or 5, characterized in that the processor (130, 266, 270) selects an alternative replacement frame to provide a minimum weighted error. 7. The processor of claim 6 wherein the processor (130, 266, 270) determines the minimum weighted error by evaluating a weighted segmental signal-to-noise (SEGSNR) or an average weighted SNR for each of the buffered frames. A speech communication unit, characterized in that. 8. The speech communication unit of claim 6 or 7, wherein the processor (130, 266, 270) determines a minimum weighted error of the subset of speech encoding parameters. 9. The processor (130) of claim 6, 7, or 8, wherein the processor (130, 266) actually searches around the pitch period of the buffered speech frames and selects the frame that exhibits the highest SEGSNR value. A speech communication unit, characterized in that. 10. Speech communication unit according to any one of the preceding claims, characterized in that the alternative speech frame (320) is referenced to the current speech frame only when transitions occur and speech is not fixed. Speech decoder 132 according to any one of the preceding claims, wherein the speech decoder 132 receives a plurality of speech frames on the transmission path 281 and one or more alternative speech frames 320 on the virtual transmission path 282. Wherein the one or more references relate to an alternative speech frame 320 within the number of speech frames received on the transmission path 281 to be used as a replacement frame when the frame is received in error. , Speech communication unit. 12. The method according to claim 11, wherein if the alternative speech frame 420 is received in error, frame 430 is selected as the alternative frame for the alternative frame 420 received in error and received in error. Speech communication unit, characterized in that it is used for the replacement of the current speech frame 410 received in error as well as the alternative speech frame. In the method for performing bad-frame error mitigation in the voice communication unit 100, The method includes transmitting, by the speech encoder 134 of the speech communication unit 100, a plurality of speech frames onto the transmission path 281 to a speech decoder, the method comprising: Transmitting on the virtual transmission path 282 one or more references to a plurality of speech frames transmitted on the transmission path 281, wherein the one or more references are replacement frames when the frame is received in error. And an alternative speech frame within the transmitted number of speech frames transmitted on the transmission path (281) to be used as. Speech communication unit (100) for performing the method steps according to claim 13. A wireless communication system supporting the use of the transmission path (281) and the virtual transmission path (282) according to any one of the preceding claims.