EP1116355A1 - Method and device for estimating the transmission quality of a digital communication signal - Google Patents

Abstract

Disclosed is a method for estimating the transmission quality of a digital communication signal. According to the inventive method, a variable is determined for each bit of information that is transmitted by a transmitter to a receiver, whereby said variable acts as a measure for the reliability of correct detection thereof and is subjected to low pass filtering. A device (1) for carrying out said method can be connected to an equalizer (3). Said device comprises a low pass filter (5) that filters the probability value and delivers a representative signal for the estimation of transmission quality.

Description

description

Method and device for / estimating the transmission quality of a digital message signal

The invention relates to a method and a device for estimating the transmission quality of a digital message signal, in which a value of the bit is assigned to bits transmitted by a transmitter at the receiver and a measure of the reliability of the correctness of the assignment is determined.

Such a method or such a device is used in particular to estimate the transmission quality in the context of a mobile radio system and to adapt a transmission mode used to the available transmission quality.

A voice coder / decoder (codec) which is to carry out such an adaptation is currently being used as the next generation at ETSI following the standardization of the GSM Enhanced Full Rate (EFR) voice codec in 1996 under the name Adaptive Multirate (AMR) voice codec SMG11 standardized. The main goals of the AMR codec are to achieve fixed network quality of the voice under different channel conditions and to ensure optimal distribution of the channel capacity. The codec is said to work under good channel conditions and / or in highly utilized cells in the half-rate (HR) channel. Under poor channel conditions, it should switch dynamically to the full rate (FR) channel using the GSM intra-cell handover and vice versa. Within a channel mode (FR or HR) there are several code modes available for different speech and channel coding rates, which should also be varied according to the channel quality (rate adaptation). Thus, taking the changing channel conditions into account, transmission should be achieved with the best possible quality. A sufficiently precise estimate of the channel quality plays a decisive role in the selection of the modes used for a transmission (i.e. when switching between channel modes FR and HR and / or between code modes), and therefore also in the entire AMR concept. Ideally, the voice quality perceived by a user should serve as a criterion for selecting a mode. It is therefore necessary to define a metric that allows such an a priori subjective quality to be measured objectively. Options for deriving such a metric of the channel quality are burst-wise RxLev, RxQual in the GSM system, DTX activation, frequency hopping activation, bit-wise or burst-wise channel state information CSI (Channel State Information) from the equalizer, a residual error rate (residual error rate) ) of the channel decoder, bad frame indicator (BFI), error concealment in the channel or speech decoder etc.

The present invention is based on an estimate of the transmission quality on the basis of channel status information (CSI), such as is supplied in the form of soft bits by an equalizer of a conventional mobile radio receiver. Such soft bits each correspond to one bit of the message signal transmitted by radio and comprise a given number of bits, for example 8 or 16. The soft bit can be a signed integer with values between -2 ^{1 "1} and 2 ^1_1 -l, i = for example 8 or 16, and provides a measure of the security with which a bit of the message signal has been recognized in the equalizer. For example, a value -2 ^{1 "1} of the soft bit denotes the reliable detection of a bit" -1 "of the Message ^signal , the value 2 ^1_1 -1 the reliable detection of the value “+ l ^λ , the value -1 being logically assigned to ONE and +1 logically ZERO. Intermediate values in each case correspond to different reliably identifications. The sign (MSB) of the soft bit includes Holds the equalizer's decision as to whether the bit of the message signal sent was +1 or -1. The amount of the soft bit indicates how safe that decision was, that is it is a measure of the reliability that the assignment of the MSB to the transmitted bit is correct.

These soft bits are conventionally used in the receiver to restore the transmitted message signal as true to the original as possible. The reliability measures contained therein are not suitable for estimating the transmission quality of a channel. The reason for this is that the transmission quality of mobile radio channels is dependent on fluctuations in the transmission quality, which can be traced back to different causes. For example, so-called short-term fading, i.e. rapid changes in reception power within a few milliseconds, is generally caused by reflection, refraction and interference in an otherwise unchanged spatial environment.

Shadowing due to slow changes in the geographical environment, caused by the movement of the individual mobile radio subscribers, leads to long-term fading, in which the average reception power changes over a period of a few seconds. The effects of short time fading on the transmission quality can be reduced in a simple manner by interleaving data blocks. Short-term deteriorations in the received signal have a strong impact on the recognition reliability of the equalizer, but, as long as they can be intercepted by interleaving, do not necessarily lead to a deterioration in the transmitted voice quality and should therefore not be taken into account when estimating them.

According to the present invention, a simple way of achieving the goal of a simple and quick estimate of the transmission quality is low-pass filtering of reliability values of a transmitted sequence of bits,

These reliability values are preferably obtained from the soft bits in that the amount of the soft bit assumed to be a signed integer is obtained. It is further preferred that the low-pass filtering is preceded by averaging over the reliability values of a given first number n of transmitted bits, in which the n bits with the lowest reliability of the assignment are selected from a given second number N of bits and the average value over the reliability values of these n bits is formed. The reason for this measure is that even if the transmission quality is poor, the equalizer often still delivers or allocates a very large number of bits with very high reliability, so that when averaging over the reliability values of all transmitted bits, the mean value obtained represents only a very insensitive measure of the transmission quality would.

The numbers n, N are preferably in a ratio of 5n <N <20n, preferably lOn≡N. A burst of a message signal transmitted according to the AMR convention comprises N = 114 bits. The n = 10 most uncertain are selected from these and used for averaging.

The low-pass filtering is preferably carried out with incomplete suppression in the stop band above a few Hz. For example, an Equipple FIR filter is suitable for this. Incomplete suppression makes it possible to react to abrupt, permanent changes in the transmission quality faster than would be the case with filtering with complete suppression.

The low-pass filtered signal is preferably compared with at least one threshold in order to obtain a comparison result which is used as a control signal for switching between different transmission modes of the message signal. In order to prevent rapid switching back and forth between transmission modes when the transmission quality fluctuates in a limit range, it is advisable to use a hysteresis when switching between different transmission modes. introduction modes. For this purpose, two different transmission modes can be assigned two thresholds in such a way that a switch is made from a first of the two transmission modes to the second if the lower of the two thresholds is undershot, and a switch from the second to the first transmission mode if the higher of the two thresholds is exceeded. If the different transmission modes have different data rates, the number N of bits from which the most unreliable soft bits are selected is preferably predetermined for each transmission mode in proportion to its data rate. This ensures that the speed at which a change in the transmission quality can be reacted to is the same for the different transmission modes, regardless of their data rate.

Further features and advantages of the invention result from the following description of exemplary embodiments with reference to the figures. Show it:

Figure 1 is a block diagram of a base station of a telecommunications system with mobile terminals, which includes a device for estimating the transmission quality according to the present invention;

FIG. 2 shows a block diagram of a mobile terminal which is equipped with a device according to the invention and communicates with the base station from FIG. 1;

FIG. 3 shows a measured course of long-term fading in the course of a message signal;

FIG. 4 shows the result of an estimation of the reception quality for the same message signal when averaged over the ten bits with the lowest reliability value within a burst; FIG. 5 shows the result when averaging over all bits of a burst;

FIG. 6 shows impulse response and frequency response of a low-pass filter of a device according to the invention; and

FIG. 7 illustrates the conversion of an estimate of the transmission quality of a message signal into a control signal for switching between different types of transmission.

FIG. 1 shows a highly schematic section of a base station for a telecommunications system that uses a device 1 to estimate the transmission quality of a digital message signal. The base station receives the digital message signal via an antenna 2. An equalizer 3 connected to the antenna 2 supplies a soft bit for each bit received by the antenna, which has a width of 8 bits, for example.

The output signal of the equalizer is fed to processing circuits for reconstructing the transmitted message signal, which are not shown in the figure. The output of the equalizer 3 is also connected to an input of a CSI generator 4 of the estimation device 1. The CSI generator 4 estimates the short term fading of the transmission channel, whereby it determines the transmission quality of each individual burst of the message signal. Depending on the transmission mode of the message signal, it contains a different number of bursts per speech frame. With full rate transmission, a speech frame comprises four bursts, with half rate transmission two.

Each burst is processed for an equalizer with a resolution of 8 bits in accordance with the C program code specified below. C Program code

The sign of each soft bit always coincides with the presumed value of the bit received, and the amount is a numerical value between 0 and 127, which is a measure of the

Reliability of the decision about the sign including tet. An amount of 0 stands for a very uncertain decision and 127 for a very safe decision.

For the 2 ⁷ = 128 possible different values of the reliability information, a temporary data field "sort" of size 128 is created and initialized with 0. In a first loop, "burst [n]", 0 <n <114 is given for the individual soft bits , a measure of the probability that the sign of the soft bit matches the corresponding bit of the transmitted message signal is first obtained by forming the amount, and the number of bits within the burst with a certain reliability value is determined and corresponding to this value in the “sort ". The index of the field represents the reliability and the content of the field the number of bits present in the burst with this reliability. For example," sort [10] = 12 "means that there are 12 bits with a reliability of 10 gives. In a second loop, starting from index 0 with the lowest reliability, the reliability values of the 10 least reliable bits are added up.

Division of the sum obtained by the number of bits added together gives a first mean.

The CSI generator 4 also carries out a second averaging, in which the above-mentioned averages over the 10 bits with the lowest reliability value of a burst are added for a number K of bursts and divided by K. The number K is 2 for a shark rate transmission and 4 for a full rate transmission. It corresponds to the number of bursts per frame, i.e. it is proportional to the data rate of the transmission mode. The dependency of the number of bursts taken into account on the transmission mode ensures that estimates of the transmission quality are available through the second averaging with a fixed repetition rate that is independent of the transmission rate. The output signal of the CSI generator 4 obtained by these averaging is approximately proportional to the short term fading of the mobile radio channel on which the message signal is transmitted. The resulting strong fluctuations in the output signal of the CSI generator 4 are suppressed with the aid of a low-pass filter 5. The reason for using the low-pass filter 5 instead of averaging over a larger time interval is that simple averaging over several frames would not lead to a satisfactory result, since short-term strong disturbances would further lead to a considerable decrease in the estimated transmission quality, which a change in the transmission mode might appear necessary, even if the decrease takes so little time that it can be compensated for by interleaving. An unweighted averaging is therefore a poor low-pass filter. Therefore, in the estimation device 1, the low-pass filter 5 with the following specifications is connected to the output of the CSI generator 4: -Filter type: FIR Equiripple low-pass filter (constant blocking range)

-Filter order: 28 -sampling rate: 50 Hz -passband: 0.2 Hz -blocking range: 1.8 Hz at 20 dB attenuation

Figure 6 shows in part A the transfer function h (t) of such a filter, part B shows the frequency response 201og (| H (2πf) | in decibels as a function of frequency f in Hz. In principle, other options for low-pass filtering are also conceivable, such as For example Butterworth, Tschebyscheff, IIR filters etc. or a weighted averaging, the weight of a soft bit decreasing with age.

FIG. 3 shows an example of the measured course of the long-term fading of a real message signal over 2000 frames, corresponding to a time period of 40 seconds (transmission rate 50 frames per second). The signal-to-noise ratio C / (I + N) is plotted in decibels on the abscissa.

FIG. 4 shows the estimate of the reception quality of the message signal provided by the low-pass filter 5 with the fading behavior shown in FIG. 3. The numerical values of the output signal of the low-pass filter 5, which can be between 0 and 127 (for 8-bit wide soft bits), are plotted on the abscissa. As can be seen, the times of the occurrence of the extremes of the signal quality from FIG. 3 and the estimation from FIG. 4 are in excellent agreement with approximately 700, 1070 and 1490 frames. The amplitude of the deflections in the estimate from FIG. 4 also agrees well with the course shown in FIG.

For comparison, FIG. 5 shows the result of an estimate in which all 114 soft bits of a burst were taken into account, and not only the ten with the lowest reliability value, as in the case of FIG. 4. The position of the extremes still agrees well with that of the extremes Figure 3 matches, but the amplitude of the rashes is reduced to about half. At 760 frames, the estimate shows a minimum to which no minimum corresponds to the measured fading curve from FIG. 3. The reliability of the estimate is therefore lower overall than in the case of FIG. 4.

As can be seen, the course of the measurement curve from FIG. 3 can be reproduced well by selecting and averaging the n = 10 bits with the lowest reliability value from a burst of N = 114 bits. It is obvious that depending on the conditions of use, the quality of the equalizer 3 or other factors, a different value for the number n of the selected bits can result in a better match of the estimate with a measured quality curve. It is assumed that a ratio of 5n <N <20n will be satisfied in the practically relevant cases. The output signal of the low-pass filter 5 is present at the input of a so-called metric generator 6. This metric generator 6 is a further developed comparator which compares the filter output signal with a plurality of thresholds and generates a control signal of 2 bits width depending on the comparison result. Horizontal lines A, B, C corresponding to the thresholds are shown in FIG. 7 above a curve which corresponds to the curve from FIG. If the output signal L _{filt of} the low-pass filter 5 is greater than the threshold B, so the transmission quality is very good, the control signal has the binary value 10. With a good channel quality with B> L _filt > A, it has the binary value 11, with a poor channel quality with A> L _filters > C the value 01 and with a very poor channel quality L _fi ι _t > 10 the value 00. As you can see, only one bit of the control signal changes when the filter output signal L _fi ι _t one of the Thresholds crossed; that is, the control signal is gray-coded.

The thresholds A, B, C are freely selectable and each indicate the limits at which the transmission mode is to be switched. They have the following meaning:

Threshold A: switching from transmission mode with the highest speech rate to transmission mode with medium speech rate when the threshold is undershot,

Threshold B: switching from the transmission mode with medium speech rate to that with the highest speech rate when the threshold is exceeded; and threshold C: versa switching from the average voice rate to the transmission mode with the lowest rate speech and vice ^¬.

By choosing a higher value for threshold B than for threshold A, a hysteresis is brought about for the switching process, that is to say the channel quality must be better for switching from the medium to the highest rate than when switching from the highest to the medium rate. This prevents constant switching between these two transmission modes if the channel quality fluctuates in the region of the thresholds A, B.

The control signal is present at a first input of a control unit 7. The control unit 7 evaluates the control signal and effects the rate adjustment for the transmission from the mobile terminal to the base station (uplink). For this purpose, it transmits a requested uplink rate (UL_REQ_Rate) inband, that is, together with the voice bits, to the mobile device. The mobile terminal, on the other hand, transmits the transmitted uplink rate as UL_RATE and the control signal to the base station.

Figure 2 shows a highly schematic block diagram of a mobile terminal that can work with the base station of Figure 1. Like the base station, it comprises an equalizer 3, which supplies soft bits to an estimation device 1 on the basis of message signals received via an antenna 2 and which, like that from FIG. 1, comprises a CSI generator 4, a low-pass filter 5 and a metric generator 6. The control signal generated by the metric generator 6 is transmitted via an antenna 8 to the control unit 7 of the base station, which, as stated above, adapts the downlink transmission mode as a function of the control signal supplied by the mobile terminal.

The control unit 7 evaluates the control signal received by the mobile terminal via the antenna 2 in the same way as that supplied by the metric generator 6 to the base station.

The conversion of the signal L _{£ i} ιt into a control signal of 2 bits width is necessary because the control unit 7. to control the rate adjustment of the downlink from the base station to the mobile terminal, information about the quality at all times des Downlinks needed, which you must be supplied by the mobile device. Very few bits are available for the transmission of this information. Transferring only the most significant bits of the filter output signal L _{fi t} would therefore result in coarse quantization. A transmission of a finer quantized or complete filter output signal, on the other hand, would have to be divided into several frames, which would, however, lead to a significant increase in the switching delay. The two-bit wide control signal of the metric generator 6, however, can in each

Speech frames are transmitted to the base station so that it can redefine the transmission mode after each speech frame.

This evaluation of the control signal in the control unit 7 takes place for uplink and downlink transmission in the same way as follows: The control signal values dual 10, 11, 01 and 00 are each assigned numerical values 3, 2, 1 and 0, which are monotonous with the transmission quality to change. The current numerical value and the last seven numerical values (ie the results of the estimation of the transmission quality for the last eight frames) are added up, and depending on the sum, a transmission mode is selected which specifies a voice transmission rate. In the case of the downlink, this is used for sending and in the case of the uplink it is sent to the mobile terminal as a command for setting an uplink rate.

Successive numerical values may only change by one level, i.e. for example a numerical value of 3 can only be followed by the numerical value 3 or 2 again. Accordingly, the transmission rate defined as a function of this can only change by one stage between two frames. This can be used as a priori information in order to minimize transmission errors and thus very disturbing speech module errors. Instead of the centralized decision about the transmission modes to be used for uplink and downlink described here by the control unit of the base station, a modification is also conceivable in which the mobile terminal independently * decides on the transmission mode to be used for uplink and / or downlink and sends corresponding setting commands to the base station.

Claims

claims 1. A method for estimating the transmission quality of a digital message signal, in which a value of the bit is assigned to a bit transmitted by a transmitter at the receiver and a measure of the reliability of the assignment is determined, characterized in that a measure of the transmission quality by low-pass filtering the Reliability values of a transmitted sequence of bits are obtained. 2. The method according to claim 1, characterized in that the value and the measure of its reliability are determined bit by bit and combined into a uniform data word (soft bit) and the low-pass filtering is carried out on the amounts of the soft bits. 3. The method according to claim 1 or 2, characterized in that the low-pass filtering is preceded by averaging over the probability values of a given first number n of transmitted bits, and the low-pass filtering is carried out over the mean values obtained 4. The method according to claim 3, characterized in that the n bits with the lowest probability of the correctness of the assignment are selected from a given second number N of bits and the mean value is formed over the probabilities of these n bits. 5. The method according to claim 4, characterized in that for each possible value of the probability, the number of those bits is determined among the N bits that have the value. 6. The method according to claim 4 or 5, since you rch characterized in that 5n <N <20n and preferably lOn≡N. 7. The method according to claim 4, 5 or 6, characterized in that the N bits each form an organizational unit of the message signal transmitted between the transmitter and receiver. 8. The method according to any one of the preceding claims, characterized in that the low-pass filtering is carried out with incomplete suppression in the restricted area. 9. The method according to any one of the preceding claims, characterized in that the low-pass filtered measure is compared with at least one threshold (A, B, C) in order to obtain a comparison result which is used as a control signal for switching between different transmission modes of the message signal . 10. The method according to claim 9, characterized in that two different transmission modes are assigned two thresholds (A, B) in such a way that a switch is made from a first of the two transmission modes to the second if the lower of the two thresholds (A) is undershot, and switching from the second to the first transmission mode when the higher (B) of the two thresholds is exceeded. 11. The method according to claim 9 or 10, as far as referred back to claim 4, characterized in that the different transmission modes have different data rates, and that the second number N is predetermined for each transmission mode proportional to the data rate. 12. Device for estimating the transmission quality of a digital message signal, for connection to the output of an equalizer (3) of a receiver for the message signal, the device (1) for the bits transmitted by a transmitter and equalizer for bits transmitted by a transmitter (3) receives assigned bit values and a measure of the reliability of the assignment of the transmitted bits, characterized in that the device (1) comprises a low-pass filter (5), which for smoothing out rapid fluctuations in the reliability measure of a transmitted sequence of bits Estimation of the transmission quality provides a representative signal. 13. The apparatus according to claim 12, characterized in that it comprises a computing circuit (4) for calculating the mean value of the reliability measures of a given number n of transmitted bits. 14. The apparatus according to claim 13, characterized in that the computing circuit (4) comprises means for selecting the n bits with the lowest degree of reliability from a set of N bits (N> n). 15. The apparatus according to claim 14, characterized in that the measure of reliability is a digital value of i bit width, and that the means for selecting (4) 2 ¹ memory locations for storing count values of the frequencies of occurrence of the representable probability values include. 16. The apparatus according to claim 15, characterized in that the N bits form an organizational unit of the message signal transmitted between a transmitter and the receiver. 17. The device according to one of claims 12 to 16, characterized in that the low-pass filter (5) has an incomplete suppression in the blocked region. 18. Device according to one of claims 12 to 17, characterized in that the low-pass filter (5) is an Equiripple-FIR low-pass filter. 19. Device according to one of claims 12 to 18, characterized by a metric generator (6) which receives the output signal of the low-pass filter (5), compares it with at least one threshold (A, B, C) and provides an output signal which defines a transmission mode to be used depending on the result of the comparison. 20. The apparatus according to claim 19, characterized in that the metric generator (6) compares the output signal of the low-pass filter (5) with two thresholds (A, B) and switches the control signal from a first state to a second when the falls below the lower (A) of the two thresholds, and switches from the second state to the first if the higher (B) of the two thresholds is exceeded. 21. A mobile terminal for a mobile radio system, characterized in that it comprises a device (1) according to one of claims 12 to 20, and in that the terminal is set up to be a representative of the device (1) provided for the estimation of the transmission quality Transfer control signal to a base station. 22. Terminal according to claim 21, as far as related to claim 19 or 20, characterized in that the control signal transmitted to the base station is the output signal of the metric generator (6). 23. Base station for a mobile radio system, characterized in that it comprises a device (1) according to one of Claims 12 to 20 and comprises a control unit (7) which, depending on a control signal representative of the estimation of the transmission quality, is used for the transmission between the base station and assigned mobile terminals used transmission mode determined. 24. Base station according to claim 23, characterized in that the control unit (7) is set up to determine the transmission mode used on the basis of a control signal transmitted by the mobile terminal.