EP1390947B1 - Method for signal reception - Google Patents

Description

The invention relates to a method for signal reception in one digital communication system using a background signal generator from received characteristic data, which one Characterize transmitter-side background signal, an output background signal is generated. Beyond concerns the invention a method for voice transmission in digital communication systems in which to transmit a Background signal at least in the speech pauses with one such a method is used, as well as a corresponding Background signal generator with which such Procedure can be carried out.

In digital communication systems, i.e. H. in communication systems, where the signal to be transmitted is digitized is transmitted in digital form from the transmitter to the receiver , for example in mobile radio systems, it happens for optimal and efficient management of the whole Systems rely on the number of bits to transmit one to keep certain information as low as possible. Such Bit reduction is usually done by a certain type source code reached. In the case of a voice transmission one speaks of a speech coding. In speech coding the speech or noise to be transmitted is divided into sections constant length, divided into so-called frames. The length or duration of such frames is usually 20 ms. A frame then becomes dependent on the speech coding method converted into a certain number of bits. When using the EFR speech codecs (enhanced full rate) in the widely used today Cellular GSM standard, for example, will be a reduction achieved from 128 or 64 kbit / s to 12.2 kbit / s.

Another way to relieve the transmission channels or more effective to use, is before a transfer between essential information and immaterial Information, for example between active language and Background noise, differentiate and only the essential Information continuously at a higher bit rate to transfer and the less essential information only in to transmit reduced form. At one already used process, the so-called discontinuous transmission (DTX) is using a speech pause detector, in general Called VAD (Voice Activity Detector), decided on the sending side, whether a full or a reduced transmission takes place. When transmitting a background noise without simultaneous speech activity is e.g. in the GSM standard only every 8th frame transmitted from the sender to the receiver. Each frame is marked accordingly, whether it is is a speech frame or a background signal frame is. On the receiving end, then inside a speech decoder recognizes whether it is a speech frame or a background signal frame. at Receiving a background signal frame then becomes the following Number of missing frames replaced again by simply the received frame is repeated in the appropriate number becomes. This is usually done in a background signal generator, for example the so-called CNG (Comfort Noise Generator), which is located next to a core speech decoder, which performs the decoding of the speech frames in the Speech decoder is located. The transfer of only part of the Frame during a background noise has the advantage that during this time the transmission channel for further calls other participants or for the transmission of others Data can be used. The completion of the missing Information about the background noise in the CNG is provided therefore, to give each listener the feeling of continuous To convey transmission so that it does not irritate is feared and an interruption of the connection.

There are also various standards with different bandwidths for speech coding. A distinction is made between the basic variants for narrowband transmission (narrow band; NB) and broadband transmission (wide band; WB). The bandwidth for narrowband transmission is, for example, 300 to 3400 Hz, which corresponds to a sampling rate (sample frequency) of 8 kHz. This is a transmission in the usual telephone quality. Broadband coding achieves better transmission quality. This is, for example, in the range from 50 to 7000 Hz, corresponding to a sampling rate of 16 kHz. This standard is already being used in part in the current fixed telephone networks for video conferences etc. In addition, further broadband encoder technologies are to be developed and standardized in the future, for example various technologies with 16, 24 or 32 kbit / s bit rate for the ITU-T system. As part of the ETSI / 3GPP standardization, a wideband AMR coder is provided, a narrowband AMR coder is already specified. In the so-called AMR (Adaptive Multi Rate) process, the ratio between the bit rate of the speech encoder and the channel encoder is varied on the transmitter side depending on the channel quality. The sum of the bit rate is selected depending on the traffic and the available capacity on the transmission channels and then kept constant. A distinction is made here e.g. B. between full rate and half rate channels.
If the AMR-WB speech decoder and the AMR-NB speech decoder are implemented on the receiver side, then it is provided there that according to the transmission between a narrowband and a broadband mode, ie the AMR-NB speech decoder and the AMR-WB- Voice decoder is switched. If necessary, such a switchover can also occur during a call. When switching from broadband to narrowband mode, the user usually perceives a significant loss in quality.

If a narrow-band signal is transmitted, there is a possibility to improve the listening comfort by decoding the received signal in an NB speech decoder (narrow-band speech decoder) and then "artificially" the entire signal within a so-called wideband extender towards the larger bandwidth expand. In principle, this can be done by any method known in the literature for artificial bandwidth expansion, which also generates signal components at frequencies which were not present in the narrowband signal or only in an attenuated manner. Such an artificial bandwidth expansion is particularly useful if it is a system with different bandwidths, for example with implementations of the AMR-NB and the AMR-WB method. These systems can then, as far as possible, transmit in broadband mode, and a switch is made to a narrowband mode if the available transmission capacity is lower. When switching to the narrowband mode, the artificial broadening on the receiver side can be used in order to at least partially provide the respective user of the device with better speech quality, which is familiar from normal broadband mode, so that the latter may not register the lower bandwidth during transmission ,
However, an artificial bandwidth expansion can also be used in a system without a broadband transmission option. A wideband extender is then used (consistently or also adjustable by the user or the network operator) in order to convey an improved voice quality impression compared to the NB transmission.

Figure 1 shows a corresponding device according to the current state of the art. Within the speech decoder is the core speech decoder (Core-SD), which the actual incoming speech signal frame S decoded and a CNG which receives the background signal frames H and a corresponding output background signal is generated. ever after whether the core SD or the CNG are active, between switched the outputs of these two facilities. The result comes from the narrowband speech decoder corresponding narrowband signal, which both the Speech as well as the background signal includes. This signal is then used in a subsequent wideband extender (in the following also WB extender abbreviated) to the desired bandwidth brought. Such a device is disclosed in the article by P. Jax and P. Vary "Wideband Extension of Telephone Speech Using a Hidden Markov Model", IEEE Speech Coding Workshop (09-2000), pages 133-135. Such an extension by means of a downstream WB extenders lead to increased listening comfort. But since every single circuit operation energy costs, this inevitably also reduces the total energy consumption of the device increased. This is particularly the case with Use of such a method in mobile devices disadvantageous as this inevitably leads to faster emptying of a battery and accordingly the operating times, until the battery needs to be recharged become.

It is an object of the present invention to provide an alternative state to this state of the art, which in simple and an inexpensive way to increase bandwidth allows the receiver side with reduced energy consumption. A background signal generator is also intended to be used be specified in this procedure.

This object is achieved by a method according to claim 1 and a background signal generator according to claim 8 solved.

According to the invention, this is done within the background signal generator an output background signal from the received characteristic data with a given bandwidth, which is greater than the bandwidth of that received by Characteristics of the characterized background signal, i. H. it will already in the background signal generator from the characteristic data the narrowband background signal the broadband Background signal generated. In the process, consequently at least for the part of the signal that is in the background signal generator, d. H. in a CNG constructed according to the invention, is processed, no downstream wideband extender needed. This will result in a significant number of circuit operations no longer needed during reception, which leads to the desired reduction in the total energy requirement leads.

Accordingly, a background signal generator according to the invention Have means for generating the output signal, which are designed so that the bandwidth of the generated output background signal is larger than that Range of characteristics characterized by the input data Background signal. Such an inventive Background signal generator is preferably part of a Speech decoder, which in any receiving device can be arranged. It is preferably the receiving device around a terminal. But it can also a receiving device within any one Communication network, for example a cellular network or a fixed telephone network.

The method or such a background signal generator can be used wherever a Output background signal generated from transmitted characteristic data will, d. h, in particular in the usually in cellular networks used methods for voice transmission, at which separate background signal frames in the speech pauses are transmitted, which are then in the receiver device within the CNG can be converted into an output background signal. The inventive method or the background signal generator are not on the current narrowband and Broadband standards are limited but can always be applied when it comes down to a narrow band transmitted signal to generate a broadband signal.

The output background signal can either be according to Editing directly to a user of the device acoustic way as background noise.

However, it can also be further processed or to get redirected.

In a preferred embodiment, the output background signal in the background signal generator using the characteristic data received from, for example, a white noise signal which creates the desired larger bandwidth having. Here, the characteristic data can preferably be LPC filter coefficients (LPC = Linear Predictive Coding) for one LPC synthesis filter included. These LPC filter coefficients contain the spectral information of the background signal. In addition, the characteristic data can also be a gain factor which include the signal energy, i. H. the volume representing the background signal. The transmission the spectral information based on the LPC filter coefficients as well as the additional transmission of a gain factor is the usual one in current standards Way to transmit characteristics for a background signal. If in the method according to the invention also these characteristics used to generate the output background signal is advantageously no change in the transmission standard necessary, d. H. it can be a device at any time work with the background signal generator according to the invention, without the respective transmitter or the transmission network must judge accordingly.

A preferred way of generating an output background signal with the desired larger bandwidth from the received LPC filter coefficients of a narrow band Filters consists of the received LPC filter coefficients into corresponding LPC filter coefficients for an LPC synthesis filter of the desired larger bandwidth convert and the converted LPC filter coefficients then feed it to an LPC synthesis filter that matches the noise signal the desired bandwidth is excited.

A background signal generator according to the invention is required on the one hand, either a noise signal generator for generation a corresponding noise signal or alternatively an input for such a noise signal. On the other hand it needs means to use the output background signal to generate the received characteristic data from the noise signal. Accordingly, these agents can be an LPC synthesis filter the desired bandwidth as well as Means to filter the LPC for a narrow band Filters as they appear in the input data are in corresponding LPC filter coefficients for the broadband Implement LPC synthesis filter.

A particularly easy way to use the LPC filter coefficients for the broadband LPC synthesis filter from the zu to determine the receiving LPC filter coefficient consists in the use of a look up table, in the "narrowband" LPC filter coefficient and "broadband" LPC filter coefficients are assigned to each other. Preferably the LPC filter coefficients are mutually sentence by sentence assigned, d. H. there are complete LPC filter coefficient sets in the table saved and assigned to each other.

This implementation is simple in that it is only a memory with an appropriate allocation table becomes. The assignment table can, for example, use a parallel Training determined using suitable estimation methods become. The memory can be a memory within act of the CNG or the speech decoder. It can but also an external one, if necessary for other purposes anyway located within the receiving device or the device Act memory on which the speech decoder or CNG Has access. The mapping table can be a single large Assignment table. But it can also be a multi-part Mapping table or by several individual mapping tables act.

A variety of other methods for implementing narrowband in broadband LPC coefficients is known from the literature.

Figur 1 eine schematische Darstellung eines schmalbandigen Sprachdecoders mit nachfolgender Bandbreitenerweiterung gemäß dem Stand der Technik,

Figur 2 eine schematische Darstellung eines schmalbandigen Sprachdecoders mit nachfolgender Bandbreitenerweiterung mit einem erfindungsgemäßen Hintergrundsignalgenerator.

Figur 3 eine schematische Darstellung eines Ausführungsbeispiels des erfindungsgemäßen Hintergrundsignalgenerators,

The invention is explained in more detail below with reference to the accompanying figures using an exemplary embodiment. Show it:

FIG. 1 shows a schematic illustration of a narrowband speech decoder with subsequent bandwidth expansion in accordance with the prior art,

Figure 2 is a schematic representation of a narrowband speech decoder with subsequent bandwidth expansion with a background signal generator according to the invention.

FIG. 3 shows a schematic illustration of an exemplary embodiment of the background signal generator according to the invention,

A comparison of Figures 1 and 2 immediately makes the difference between the inventive method and the currently usual procedures evident.

Both methods have in common that in a speech decoder (SD) 8 a CNG 1 and a narrow-band core speech decoder (Core SD) 7 are arranged. The incoming signal includes each speech frame S and - in speech pauses - background signal frame H, the corresponding for each Voice decoders are recognizable. The speech signal frames S are decoded within the core speech decoder 7 and therefrom a narrowband speech signal SS is generated, which in a subsequent wideband extender into a broadband Signal is implemented. The difference of the invention Structure to the prior art consists in the structure of the Background signal generator 1, i.e. H. in the CNG 1. At the state of the Technology according to Figure 1 becomes a conventional narrow band CNG is used, which comes from an incoming background signal frame H generates a narrowband background signal. Via a switch, which corresponds to the activity of the core speech decoder and the CNG is switched put the signal on an output line of the speech decoder, so that on this output line a continuous narrowband signal is present, which is then the wideband extender is fed. That is, in this conventional method the output background signal is generated in narrowband and then in the wideband extender accordingly extensive switching operation converted into a broadband signal. Alternatively, according to the prior art Output signal of the CNG also initially in the core speech decoder to be processed further, which in turn then outputs decoded background signal.

In the method according to the invention in accordance with FIG a broadband CNG 1 used, which immediately from an incoming background signal frame H one narrowband background signal a broadband output background signal HS generated at the output. This broadband Background signal HS no longer needs the wideband extender 10 to be fed. Accordingly, only that of Core speech decoder 7 outgoing signal SS the wideband extender 10 fed. The switch 9, depending on the activity the core speech decoder 7 or the CNG 1 between the outputs this device switches, is therefore behind the wideband extender 10 arranged.

Figure 3 shows a particularly simple and inexpensive Embodiment for a CNG 1 according to the invention, which from the background signal frame H of a narrow band Background signal is a broadband output background signal HS generated. The advantage of this CNG 1 is that that as an input signal normal, according to the current one Standard transmitted background signal frame H used can be. That is, a transmission of the currently is sufficient usual LPC filter coefficients, which the spectral information of the signal included, and the transmission of the gain VF, which represents the signal energy, out. Therefore, there are no changes to a transmission standard or something similar.

The background signal generator according to the invention is accordingly also very similar to a conventional background signal generator built and has an LPC synthesis filter 2, which one by means of the white noise RS Noise signal generator 6 is excited.

The volume of the signal is determined by the gain factor VF, which is set in an amplification device 5 the white noise signal present at the input of the LPC filter RS reinforced accordingly. This gain factor becomes VF by means of an analysis device 4 from the incoming background signal frame determined.

The main difference from a CNG of the prior art is that instead of an LPC filter the bandwidth corresponding to the incoming background signal frame H here uses an LPC filter 2 of the desired bandwidth and, accordingly, the noise signal generator 6 no noise in the bandwidth of the transmitted background signal, but a noise signal RS of the desired wider Bandwidth generated.

In addition, those in the background signal frame H transmitted LPC filter coefficient directly to the LPC filter 2 fed, but these LPC filter coefficients are initially in a conversion device 3 from the background signal frame H determined and in LPC filter coefficients FK the desired larger bandwidth. This conversion takes place, for example, with the help of one in one Memory 3a stored allocation table LUT. In this LUT assignment table are the various LPC filter coefficient sets, which is the background signal of the broadcast Correspond to bandwidth, each LPC filter coefficient sets the desired bandwidth of the broadband output background signal HS assigned.

In the method according to the invention, the bandwidth widening directly integrated into the CNG 1, creating a efficient method of increasing the bandwidth for the periods the language breaks are possible. Because the sum of the pauses in speech a fairly high proportion during an overall conversation can assume is overall of a considerably smaller Energy consumption of such a receiver. On Another advantage of this method is that the Handling the actual transmitted speech signal, d. H. the treatment of the transmitted voice frame S during the Periods of language activities, not affected, so that here with conventional devices according to the known Standards can be continued.

Claims (14)

1. A method for receiving signals in a digital communication system, wherein an output background signal (HS) is generated in a background signal generator (1) from received characteristic data (H) which characterizes a background signal on the transmitter side,
   characterized in that
   an output background signal (HS) having a predetermined bandwidth is generated within the background signal generator (1) from the received characteristic data (H), said bandwidth being wider than a bandwidth of the background signal which is characterized by the received characteristic data (H).
2. The method as claimed in Claim 1,
   characterized in that
   the output background signal (HS) is generated in the background signal generator (1), by means of the received characteristic data (H), from a noise signal (RS) which has the desired wider bandwidth.
3. The method as claimed in Claim 1 or 2,
   characterized in that
   the characteristic data contains LPC filter coefficients (FK) for an LPC synthesis filter and/or an amplification factor (VF).
4. The method as claimed in Claim 3,
   characterized in that
   corresponding LPC filter coefficients (FK) are determined for an LPC synthesis filter (2) of the desired wider bandwidth on the basis of the received LPC filter coefficients, said LPC synthesis filter being activated by the noise signal (RS) of the desired bandwidth.
5. The method as claimed in Claim 4,
   characterized in that
   the ascertainment of the LPC filter coefficients (FK) for the wider-band LPC synthesis filter, from the received LPC filter coefficients, is done by means of a lookup table (LUT) in which LPC filter coefficients for an LPC synthesis filter of the bandwidth of the background signal on the transmitter side and LPC filter coefficients (FK) for the wider-band LPC synthesis filter (2) are assigned to each other.
6. A method for transmitting signals in digital communication systems, wherein for transmitting a background signal at least in the voice pauses, characteristic data (H) which characterizes the relevant background signal is transmitted to a receiver and an output background signal (HS) is generated on the receiver side,
   characterized in that
   the output background signal (HS) is generated according to a method as claimed in one of the Claims 1 to 5.
7. The method as claimed in one of the Claims 1 to 6,
   characterized in that
   the digital communication system comprises a mobile radio system.
8. A background signal generator (1) having means (2, 3, 4, 5, 6) for generating a corresponding output background signal (HS) from characteristic data (H) on the input side, said characteristic data characterizing a background signal on the transmitter side,
   characterized in that
   the means (2, 3, 4, 5, 6) for generating the output background signal (HS) are configured in such a way that the bandwidth of the generated output background signal (HS) is wider than a bandwidth of the background signal which is characterized by the characteristic data (H) on the input side.
9. The background signal generator as claimed in Claim 8,
   characterized by
   a noise signal generator (6) for generating a noise signal (RS) which has the desired wider bandwidth, and/or an input for such a noise signal, and means (2, 3, 5) for generating the output background signal (HS) from the noise signal (RS) by means of the received characteristic data (H).
10. The background signal generator as claimed in Claim 9,
    characterized by
    an LPC synthesis filter (2) of the desired bandwidth and means (3) for converting LPC filter coefficients which are contained in the characteristic data (H) on the input side into corresponding LPC filter coefficients (FK) for the LPC synthesis filter (2).
11. The background signal generator as claimed in Claim 10,
    characterized by
    a memory (3a) including a lookup table (LUT) in which LPC filter coefficients of the bandwidth of the background signal on the transmitter side and LPC filter coefficients (FK) for the wider-band LPC synthesis filter (2) are assigned to each other.
12. A voice decoder (8) including a background signal generator (1) as claimed in one of the preceding Claims 8 to 11.
13. A receiving device including a background signal generator as claimed in one of the preceding Claims 8 to 11 or including a voice decoder as claimed in Claim 12.
14. A terminal including a receiving device as claimed in Claim 13.

Images