EP2245622B1 - Method and means for decoding background noise information - Google Patents

Description

The invention relates to methods and means for decoding background noise information in speech signal coding methods.

For telephone calls, a bandwidth restriction has been provided for analog voice transmission since the beginning of telecommunications. Voice transmission takes place on: a restricted frequency range from 300 Hz to 3400 Hz.

Such a limited frequency range is also provided in many speech signal coding methods for today's digital telecommunications. Prior to a coding process, a Bandebreiteabegrezung the analog signal is performed for this purpose. For coding and decoding, a codec is used which, due to the described bandwidth limitation in the frequency range between 300 Hz and 3400 Hz, is also referred to below as narrow-band speech codec (Narrow Band Speech Codec). The term codec is understood to mean both the coding rule for the digital coding of audio signals and the decoding rule for the decoding of data with the aim of reconstructing the audio signal.

For example, a narrowband speech codec is known from ITU-T Recommendation G.729. By means of the encoding rule described there, a transmission of a narrowband speech signal with a data rate of 8 kbit / s is provided.

Furthermore, so-called broadband speech codecs (Wide Band Speech Codec) are known, which provide a coding of a in an extended frequency range to improve the impression. Such an extended frequency range lies for example between a frequency of 50 Hz and 7000 Hz. A broadband speech codec is known, for example, from the ITU-T Recommendation G.729.EV.

Usually, coding methods for broadband speech codecs are made scalable. With a scalability. Here, it is meant that the transmitted encoded data exhibits various demarcated blocks containing the narrowband portion, wideband portion and / or full bandwidth of the encoded voice signal. On the one hand, such a scalable design allows for backward compatibility on the receiver side and, on the other hand, offers a simple possibility of adapting the data rate and the size of transmitted data frames in the transmission channel in the case of limited data transmission capacities in the transmission channel.

For a reduction of the data transmission rate by a codec is usually provided a compression of the data to be transmitted. Compression is achieved, for example, by coding methods, in which parameters for an excitation signal and filter parameters are determined for coding the speech data. The filter parameters and parameters specifying the excitation signal are then transmitted to the receiver. There, a synthetic speech signal is synthesized using the codec, which is as similar as possible to the original speech signal in terms of a subjective hearing impression. With the help of this method, also referred to as "analysis-by-synthesis", the determined and digitized samples (samples) are not transmitted themselves, but determined parameters which enable a receiver-side synthesis of the speech signal.

A further measure for reducing the data transmission rate is provided by a method for discontinuous transmission (Discontinuous Transmission), which is also familiar in the art under the term DTX. The fundamental goal DTX is a reduction in the data transfer rate in the event of a speech break.

For this purpose, a voice pause detection (VAD) is used on the part of the transmitter, which detects when a certain signal level falls below a speech break.

Usually the receiver is not expected to have complete silence during a speech break. On the contrary, a complete silence on the receiver side would lead to irritation or even to the suspicion of a breakdown of the connection. For this reason, methods for generating a so-called comfort noise (Comfort Noise) are turned away.

Comfort noise is noise that is synthesized to fill silence phases on the receiver's side. The comfort noise serves as a subjective impression of a continuing connection, without claiming the data transmission rate intended for the transmission of speech signals. In other words, less effort is required to code the speech data for the transmitter-side coding of the noise. For a receiver-side still perceived as realistic synthesis - i. Decoding - comfort noise transmits data at a much lower data rate. The data transmitted here are also referred to in the art as SID (Silence Insertion Description).

Problems arise in the prior art with the method of discontinuous transmission when using broadband speech codecs, such as ITU-T G.729.1, G.72.2.2 or 3GPP AMR-WB. Said scalable wideband speech codecs usually support different data transmission rates in a bandwidth range of 50 to 7000 Hz.

Possible data rates for encoding the speech information are e.g. 8, 12, 14, 16, ..., 32 kbit / s, which is e.g. are used in standard G.729.1. The data rates of 8 and 12 kbit / s are applied to narrowband signals (50Hz to 4kHz). Data rates greater than 12 kbit / s are applied to the upper frequency band of 4 to 7 kHz.

During a transfer, a change between the said data rates is possible. A sudden change from a narrowband to a broadband data rate is known to cause a disturbing effect on a human recipient. Such a transition occurs, for example, as a result of a truncation of the data stream (bitstream transduction), which is caused, for example, by the transmission network between transmitter and receiver, for example as a result of the establishment of further additional connections or due to a congestion in the transmission network. The said truncation leads to a change in the data rate and finally to a transition from a broadband to a narrowband transmission of the speech signal.

If in the encoder method for discontinuous transmission - or DTX method - applied, a saving of the data transmission rate for the transmission of the respective data frames is possible. The DTX method is used exactly when a corresponding frame is characterized as a speech break. Using the DTX method, a reduced data transfer rate on transmitted frames is achieved due to two factors. First, the encoder does not need to send all inactive frames to the decoder. Second, a transmitted SID frame occupies much fewer bits than a voice data frame.

Such a method requires involvement of a speech pause recognition (VAD) on the encoder side. With the help of a speech pause detector, the encoder on the transmitter side is informed whether a current sample containing and to encoded frame contains a speech signal or a speech break with background noise. With the help of this characterization measures are taken in the encoder which determine the perceptional characteristics of an inactive frame (Inactive Speech Frame). Such perception characteristics include, for example, the average energy and spectral and temporal characteristics.

The encoder then sends a specially marked frame, a Silence Insertion Descriptor (SID) frame, to the decoder. The decoder synthesizes comfort noise based on the information contained in a SID frame, and the decoder can determine, based on the SID frame, whether the contained noise information is narrowband or broadband information.

Changing the bit rate ("bit rate switching") between narrowband and wideband information is a common scenario for any scalable wideband speech codec. A treatment of a data rate change during a normal speech phase, i. in the absence of pauses in speech, although adequately described in the literature, a treatment on entry into a DTX phase is currently still unknown.

Therefore, there is an urgent need to provide a method of data rate switching during a DTX phase and / or entering a DTX phase to optimally respond to a change between a narrowband and wideband data rate before or during the transition to the DTX phase to react.

During a speech pause, data rate truncation is unlikely because the bitstream allocation of a SID frame requires fewer bits anyway than an active speech data frame in a "normal" codec operation, So a codec operation during a exclusive speech phase.

This leads to a possible scenario in which the data rate is changed during an active speech phase but remains in a broadband mode during pauses in speech, ie during the DTX phase. As a very disturbing part of a human receiver on the decoder side of the case is felt, in which the active speech frames are narrow-band decoded and the background noise is reproduced in speech breaks broadband.

This case occurs with high probability e.g. in situations where the speech data frames sent on the encoder side are truncated by the transmission network, but the transmission network still leaves enough capacity to transmit the wideband SID frames.

So far, no methods for changing the data rate of SID frames during a speech break are known. The existing data rate switching methods only relate to normal codec operation during an active speech phase.

The object of the invention is to specify a method for changing a data rate of SID frames during a speech pause, which results in an improved quality of the signal synthesized on the decoder.

The object is solved by the subject matter of the independent claims.

A basic idea of the invention is to carry out a determination of information about the course of the bit rate switching during an active speech phase. The scalable nature of the speech signal coding method or codec used according to the invention expresses while the possibility of the codec for bandwidth switching already off.

According to the invention, during the speech phase on the part of the decoder, information is collected on the percentage of broadband active speech frames in comparison to narrow-band active speech frames. In other words, information on the nature of the background noise is not collected until the time of a change to a speech pause, as has hitherto been suggested by the prior art. A higher percentage of broadband active speech frames indicates that a broadband use is preferred by the codec and therefore there is a need to broadly synthesize noise information during a DTX phase, i. to decode. On the other hand, if a low percentage is detected, narrow-band noise is generated on the decoder's side when entering a DTX phase, even if the received SID frames would allow for synthesizing - i.e. decoding - a wideband noise.

With the inventive method presented here, the object of the invention to provide a method for changing a data rate of SID frames during a speech break more than solved. The change to be made between noise information with different data rate according to the object is refined according to the inventive solution presented here into a determination of a proportion of noise information with different data rates. The proportion is adjustable in contrast to a change in any ratio between noise information with different data rate.

The adjustability or adaptability of the noise signal quality to the previously raised speech signal quality (narrowband / broadband) results for the entire signal, ie noise and speech signal, on the receiver side overall a significantly increased signal quality. The method according to the invention thus achieves the object of the invention to achieve an improved quality of the signal synthesized on the decoder.

Such an approach according to the inventive method proves to be fundamental for advantageous developments of the invention, which are the subject of the dependent claims.

If, according to the method of the invention, a decision is made that a noise signal having a certain quality (ie, wideband or narrowband) is synthesized during a speech pause, a case may occur in which a truncation occurs in the last few frames during an active speech phase on the part of the network the active data frame took place.

To explain, it is first assumed that the codec used preferred a broadband reproduction mode and a broadband transmission method in the past was also predominantly ensured by the transmission network. This may lead to the case that few active speech frames arrive at the receiving decoder as narrowband speech frames before first SID frames are detected there.

In this case, without additional measures, an abrupt transition from a narrowband speech signal to a wideband noise signal would occur during the first few SID frames. As important as such a transition to reinstatement to a broadband reception condition is in general, however, this transition is perceived as disturbing to the recipient.

According to one embodiment of the invention, it is therefore provided that, when entering the DTX phase, a predominantly narrow-band decoding of the background noise information first takes place, which after a settable period of time transitions into a predominantly wideband decoding. Such a transition is thus preferably quasi-continuous, with a Transition to discrete times - hence "quasi" -continuous - is set to a certain share factor.

According to one embodiment of the invention, a fast switching method is proposed, in which a quasi-continuous transition from a narrowband (share factor = 0) to a broadband (share factor = 1) noise signal quality is performed within a certain time frame of 100 ms , This transition is performed by the decoder.

• zum Zeitpunkt des Eintritts in die. DTX-Phase ein Anteilsfäktor von 0, mithin ausschließlich schmalbandiges Rauschen;
• zu einem Zeitpunkt von 20 ms nach Eintritt in die DTX-Phase ein Anteilsfaktor von 0,09525986892242;
• zu einem Zeitpunkt von 40 ms nach Eintritt in die DTX-Phase ein Anteilsfaktor von 0,19753086419753;
• zu einem Zeitpunkt von 60 ms nach Eintritt in die DTX-Phase ein Anteilsfaktor von 0,36595031245237:
• zu einem Zeitpunkt von 80 ms nach Eintritt in die DTX-Phase ein Anteilsfaktor von 0,62429507696997; und;
• zu einem Zeitpunkt von 100 ms nach Eintritt in die DTX-Phase ein Anteilsfaktor von 1, mithin ausschließlich breitbandiges Rauschen.

According to a development of the invention, the following values for the proportion factor have proved to be particularly advantageous for the subjective perception of human hearing:

• at the time of entering the. DTX phase a fraction of 0, hence only narrowband noise;
• at a time of 20 ms after entering the DTX phase, a share factor of 0.09525986892242;
• at a time of 40 ms after entering the DTX phase, a proportion factor of 0.19753086419753;
• At a time of 60 ms after entering the DTX phase, a Share Factor of 0.36595031245237:
• at a time of 80 ms after entering the DTX phase, a proportion factor of 0.62429507696997; and;
• at a time of 100 ms after entering the DTX phase, a proportion factor of 1, hence exclusively broadband noise.

According to one embodiment of the invention, it is assumed that the codec used was a narrow-band playback mode Favor and / or a broadband transmission in the past by the transmission network was not guaranteed. This may lead to the case that few active speech frames arrive at the receiving decoder as wideband speech frames before receiving first SID frames there.

According to one embodiment of the invention, it is provided that, when entering the DTX phase, a predominantly wideband decoding of the background noise information initially takes place, which after a settable period of time transitions into a predominantly narrowband decoding. Such a transition is preferably quasi-continuous analogous to the development described above, wherein a transition to discrete times is set to a certain proportion factor.

According to one embodiment of the invention, a fast switching method is proposed, in which a quasi-steady transition from a broadband (share factor = 1) to a narrowband (share factor = 0) noise signal quality is performed within a certain time frame of 100 ms , This transition is performed by the decoder.

For the quasi-steady transition from the broadband to the narrowband noise signal quality, the proportional factor is set with values as above, but in reverse order.

An embodiment with further advantages and embodiments of the invention will be explained in more detail below with reference to the drawing.

Fig. 1

eine zeitliche Darstellung einer batenrate zwischen einem Sender und einem Empfänger mit mehreren Bandbreitenumschaltungen und einem Eintritt in eine Sprechpause, wobei SID-Rahmen gesendet werden;

Fig. 2A

eine schematische Darstellung eines erstes Szenarios einer Bandbreiteriumschaltung;

Fig. 2B

eine schematische Darstellung eines zweiten Szenarios einer Bandbreitenumschaltung; und;

Fig. 3

Ein dekoderseitig ausgeführter Umschaltvorgang mit einem quasi-stetigen Übergang von einer schmalbandigen auf eine breitbandige Rauschsignalqualität.

Showing:

Fig. 1

a timing of a batenrate between a transmitter and a receiver with multiple bandwidth switches and an entry into a pause, wherein SID frames are sent;

Fig. 2A

a schematic representation of a first scenario of a Bandbreiteriumschaltung;

Fig. 2B

a schematic representation of a second scenario of bandwidth switching; and;

Fig. 3

A decoder-executed switching process with a quasi-steady transition from a narrowband to a broadband noise signal quality.

In Fig. 1 is a temporal transmission of Sprachdatenrah-5 men with a respective data rate DR and, as of a third time t3, a transmission of SID frames shown.

Before a first time t1 there is a transmission of broadband active speech frames with a data rate of 32 kbit / s. From the time t1, a switchover to a data rate of 22 kbit / s and from a second time t2 to a data rate of 12 kbit / s. A data rate of 12 kbit / s already corresponds to a narrowband speech frame.

At a third time t3, it is assumed that a transition to a DTX phase occurs due to a pause in speech on the part of the transmitter. From the third time t3, SID frames SID are thus sent in a certain time period.

From the third time t3, the previously explained situation occurs that in the past - during the time phase between the second time t2, and the third time t3 - a narrowband voice signal was transmitted, wherein from the third time t3 now a broadband noise signal appropriate SID cramping available is provided. The data rate of the SID frames is 43 bit / 20ms = 2.15 kbit / s with a length of 43 bits per SID frame and a period of 20 ms per sent SID frame.

In this situation, the case occurs that the decoder imposes an immediate, i. discontinuous, transition from a narrowband speech signal to a broadband noise signal would occur. Such an abrupt transition is extremely annoying to a human recipient.

FIG. 2A and FIG. 2B show two possible scenarios for a progression of the data rate DR over time t.

In FIG. 2A is due to limitations of the network or due to other circumstances a transmission largely narrowband, in the example of Fig. 2A with 8. kbit / s, while at a few points in time, between a first time t1 and a second time t2, exceptionally, a broadband transmission with 32 kbit / s takes place.

In FIG. 2B again, there is a reverse situation, namely a predominantly broadband transmission mode with 32 kbit / s and exceptionally short, between a fourth time t4 and a fifth time t5 takes place, narrow-band transmission.

The following is at a time t3 for the example of Fig. 2A and at a time t6 for the example of Fig. 2B assumed that entry into a DTX phase takes place.

According to the procedure according to the invention, information about the proportion of broadband active speech frames in comparison to narrowband active speech frames will be raised on the decoder side during the speech phase.

For the example of Fig. 2A Here, the percentage of broadband active speech frames is to be described as very low, while in the example of Fig. 2B a high percentage Proportion of broadband active speech frames.

When entering a DTX phase at time t3 in the example of Fig. 2A Now, a narrow-band noise is generated using the method according to the invention, although the received from the time t3 - not shown - SID frame would allow a synthesis of a broadband noise.

In the example of Fig. 2B By contrast, at a time t6 with the DTX phase beginning there, a broadband synthesizing of the noise information is preferred.

In Fig. 3 is a noise signal quality HB-SHARE over a time TIME, which is indicated in ms, plotted. The Fig. 3 In this case, a design of the noise signal in connection with a scenario according to the preceding FIG. 2B in which, based on the decoder-side determined percentage of broadband active speech frames, a need has been identified for broadband synthesizing noise information during the DTX phase.

• Zum Zeitpunkt TIME = 20 ms ein Anteil HB-SHARE von 0,09525986892242;
• Zum Zeitpunkt TIME = 40 ms ein Anteil HB-SHARE von 0,19753086419753;
• Zum Zeitpunkt TIME = 60 ms ein Anteil HB-SHARE von 0,36595031245237; und;
• Zum Zeitpunkt TIME = 80 ms ein Anteil HB-SHARE von 0,62429507696997.

The transition into the DTX phase takes place in the representation of the Fig. 3 at the time TIME of 0 ms. To make this transition from a narrow-band speech signal to a broadband noise signal quasi-continuous, which has proven to be the most favorable embodiment for a subjective hearing of a human recipient, TIME is started at this time with an exclusively narrow-band signal, ie with a proportion HB- SHARE of the wideband noise of 0. At a time of 100 ms the ready-to-use noise is 1 or 140%. For the quasi-continuous transition from an exclusively narrow-band noise signal at the time TIME = 0 ms on an exclusively broadband noise signal at the time TIME = 100 ms in practice, the following further values of the HB-SHARE fraction have been preserved at discrete times TIME:

• At the time TIME = 20 ms a proportion HB-SHARE of 0.09525986892242;
• At the time TIME = 40 ms a fraction HB-SHARE of 0.19753086419753;
• At the time TIME = 60 ms, a HB-SHARE fraction of 0.36595031245237; and;
• At the time TIME = 80 ms a share HB-SHARE of 0.62429507696997.

A further embodiment of the invention analogously provides for a transition from a wideband speech signal to a narrowband noise signal.

For this purpose, reference is first made to Fig. 2A adopted a slightly modified scenario in which, unlike in Fig. 2A represented scenario shortly before the time t3 still not shown - change to a broadband transmission at 32 kbit / s takes place. Despite this "peak", the percentage of broadband active speech frames remains very low, so that now on transition to the DTX phase, a noise signal is to be synthesized that broadband begins, however - due to the predominantly narrow-band transmission history and thus expected for the future Continuation of the narrow-band transmission character - is to be converted into a narrow-band noise signal. In order to make this transition from a broadband speech signal to a narrowband noise signal quasi-continuous, the entry into the DTX phase is started with an exclusively broadband signal, ie with a HB-SHARE component of the broadband noise of 1. At a time of 100 ms is the narrow-band noise component 0. For the quasi-continuous transition from an exclusively broadband noise signal at the time of entry into the DTX phase to an exclusively narrow-band noise signal At a time after 100 ms, the values proposed above are advantageously set in an inverse row. This would correspond to a curve mirrored at the ordinate HB-SHARE Fig. 3 ,

Claims (15)

1. A method for decoding an SID frame (SID) for transmitting background noise information using a scalable speech signal coding method with the following steps:

   Determining an amount of received broadband speech frames relative to received narrowband speech frames during a speech phase,

   Decoding the background noise information contained in an SID frame upon entrance into a DTX phase, wherein the decoding takes place according to the determined amount of primarily broadband or narrowband.
2. The method according to Claim 1, characterized in that in the case of a determination of a high amount of received broadband speech frames, upon entrance into the DTX phase a primarily broadband decoding of the background noise information takes place.
3. The method according to Claim 2, characterized in that upon entrance into the DTX phase at first a primarily narrowband decoding of the background noise information takes place which merges after an adjustable time period into a primarily broadband decoding.
4. The method according to Claim 3, characterized in that the transition into the primarily broadband decoding can be adjusted with an amount factor (HB-SHARE) which expresses a relationship between a broadband and a narrowband noise signal quality.
5. The method according to Claim 4, characterized in that the amount factor is dimensioned to zero at the time of the entrance into the DTX phase.
6. The method according to one of Claims 4 to 5, characterized in that the amount factor is dimensioned to one at a time of 100 ms after entrance into the DTX phase.
7. The method according to one of Claims 4 to 6, characterized in that the amount factor

   - is dimensioned to 0.09525986892242 at a time of 20 ms after entrance into the DTX phase;

   - is dimensioned to 0.19753086419753 at a time of 40 ms after entrance into the DTX phase;

   - is dimensioned to 0.36595031245237 at a time of 60 ms after entrance into the DTX phase; and

   - is dimensioned to 0.62429507696997 at a time of 80 ms after entrance into the DTX phase.
8. The method according to Claim 1, characterized in that in the case of a determination of a low amount of received broadband speech frames upon entrance into the DTX phase a primarily narrowband decoding of the background noise information takes place.
9. The method according to Claim 8, characterized in that upon entrance into the DTX phase at first a primarily broadband decoding of the background noise information takes place which merges after an adjustable time period into a primarily narrowband decoding.
10. The method according to Claim 9, characterized in that the transition into the primarily narrowband decoding can be adjusted with the amount factor (HB-SHARE), that expresses a relationship between a broadband and a narrowband noise signal quality.
11. The method according to Claim 10, characterized in that the amount factor is dimensioned to one at the time of the entrance into the DTX phase.
12. The method according to one of Claims 10 to 11, characterized in that the amount factor is dimensioned to zero at a time period of 100 ms after entrance into the DTX phase.
13. The method according to one of Claims 10 to 12, characterized in that the amount factor

    - is dimensioned to 0. 62429507696997 at a time of 20 ms after entrance into the DTX phase;

    - is dimensioned to 0.36595031245237 at a time of 40 ms after entrance into the DTX phase;

    - is dimensioned to 0.19753086419753 at a time of 60 milliseconds after entrance into the DTX phase; and

    - is dimensioned to 0. 9525986892242 at a time of 80 ms after entrance into the DTX phase.
14. A codec with means for carrying out the method according to one of Claims 1 to 13.
15. The codec according to Claim 14, characterized by an implementation in the known ITU-T standard G.729.1.

Images