MXPA99010617A - Speech quality measurement in mobile telecommunication networks based on radio link parameters - Google Patents

Abstract

A method and system for measuring the speech quality in a mobile cellular telecommunications network using available radio link parameters is disclosed herein. In a preferred embodiment, the method includes receiving a set of radio link parameters, as defined in a standard or otherwise available, such as the BER, FER, RxLev, handover statistics, soft information, and speech energy. Temporal information is obtained from the radio link parameters to create a set of temporal parameters which can be statistically analyzed, for example, for the maximum and minimum, mean, standard deviation, and autocorrelation values for a time interval. The temporal parameters are combined to yield a set of correlated parameters that are more closely related to the speech quality. An estimator then uses the correlated parameters to calculate an estimate for the speech quality. The method of the present invention takes advantage of temporal information and correlated relationships from the transmitted parameters. Furthermore, the method is inherently simple and reliable as compared to prior art methods. Still further, the technique provides a method and allows operator to monitor quality conditions throughout the network.

Description

APPARATUS FOR MEASURING VOICE QUALITY IN MOBILE TELECOMMUNICATION NETWORKS BASED ON LINK PARARS RADIO FIELD OF THE INVENTION The present invention relates, in general terms, to the measurement of voice quality in wireless telecommunications systems, and refers more specifically to a od for measuring voice quality. using radio link parars. BACKGROUND OF THE INVENTION In the wireless telecommunications industry, cellular service providers "have a vested interest in providing reliable and high-quality services to their customers in today's highly competitive environment, for example, reliability issues such as, for example, Interrupted calls and quality issues such as fading, multipath interference, as well as crosstalk interference are concerns that are constantly faced by cellular operators Another issue of great interest to operators is the improvement of perceived quality Therefore, it is desirable for operators to be able to determine which areas in the network are having quality problems.In the past there were numerous ods for measuring voice quality in cellular networks. A commonly used od includes testing a cell network r through the transmission of known signals and the comparison of the received signals with a database of predefined signals in order to determine a quality estimate. The term signal is used here to refer to perceptible sounds in the human audio frequency range that include voice and tones. This od is illustrated in Figure 1. A database of known signals 2 is represented, where predetermined signals are sent through a system in test 4. The system in test 4 represents all the working components of a cellular network that includes a mobile switching center (MSC), a radio base station (RBS), all communication links, and the air interface. Once the transmitted signals are received, a second signal database 6 containing the original signal patterns is compared to the signals received in step 8. An estimate is then calculated as to the quality of the signal received by the network . In digital systems, the conversion of analog voice signals into digital signals requires much greater bandwidth for transmission than is desirable. Limitations on bandwidth in wireless telecommunication systems have resulted in the need for low bit rate speech coders that work by reducing the number of bits that are needed to transmit while maintaining both quality and understanding. In general, it is desirable to transmit at lower bit rates but the quality tends to decrease with a decrease in bit rates. The voice coders used in these requests work by voice coding while removing built-in redundancies during voice production. Typically, speech coders obtain their low bit rates by modeling the production of the human voice in order to obtain a more efficient representation of the speech signal. The original voice signal can be synthesized using several estimated filter parars. Since many of the prior art test ods include the use of audio tones in the test procedure, they do not lend themselves well to testing with digital systems. This is due to the fact that voice coders are modeled according to the voice output and __ are not optimal for tones, therefore it is likely to find errors in the regeneration of. tones Another source of potential problems with the figure od 1 when using voice signals is in the comparison and estimation step 8. The voice database 2 contains a limited number of repetitive predetermined phrases (for example 6-8 sentences) which are representative of voice patterns typically performed in a mobile network. The estimation portion in step 8 uses perceptual models that mimic the listening process. Models of this type are typically very complicated and difficult to formulate. This causes differences between the model and the subjective evaluation, thus causing sometimes unreliable measurements. A predominant factor that affects voice quality in digital systems is the proportion of errors in the bits (BER). Errors in the bits tend to be introduced during the transmission in the air interface. The BER is the proportion with which these bit errors are introduced in the frames transmitted. Situations in which the BER is high occur frequently during conditions of high interference of coca fieldsweak signals such as mobile displacement out of range, and fading caused by multipath interference due to obstructions such as buildings, etc. Even when attempts were made to correct these errors, an excessively high BER has a negative effect on the quality of the voice. In a Global System for Mobile Communication Network (GSM), for example, the BER and other related parameters such as, for example, the Quality of Reception (RxQual) and Reception Level (RxLev), are monitored in order to evaluate the voice quality There are limitations in the use of this method since the correlation relation and the temporal information that can be obtained from the parameters is not used to obtain parameters that are more closely related to the voice quality. For example, the extraction of temporary information allows the formulation of a set of relationships between the variables that can be used to measure voice quality. It is known that the quality of voice perceived by the end user is related to an average time of a phrase length at its highest resolution. The final quality is averaged over the entire conversation which means that the lowest resolution is approximately within the range of several minutes. Therefore, the use of derivative and correlated temporal parameters, which are missing in GSM, provides a clearer insight into the state of voice quality that is experienced in various situations. The RxQual parameter in the GSM system is measured every 0.5 seconds and depends inherently on the BER for each frame of 20 milliseconds. In addition, RxQual can fluctuate widely due to fading, noise or interference, which can cause quality measurements that fluctuate much faster than the perceived quality of speech. A seemingly obvious solution would be to increase the temporal resolution with a time constant in the area of 2-5 seconds. But it has been found that the relationship between the digital communication link and the voice quality does not depend only on a BER averaged over time. What is required is a method that is both simpler and more precise than the use of signal databases and takes advantage of correlation relationships and temporal information from radio link parameters. An additional objective is to offer an effective method, using available parameters, which allows operators to monitor quality conditions throughout the network. SUMMARY OF THE INVENTION In order to achieve the foregoing and other objectives in accordance with the present invention, a method and arrangement for measuring voice quality in a mobile communication network is presented herein. In a preferred embodiment, the method includes receiving a set of radio link parameters in accordance with what is defined in a standard or otherwise available, such as BER, FER (Frame Erase Rate), RxLev, statistics of transfers, as well as soft information. The radio link parameters are processed to retrieve the applicable temporal information, which is used for 1 calculate _ a set of temporary parameters. The temporal processing also includes, if necessary, the transformation of radio link parameters in the time domain to obtain more manageable forms. The transformed data can then be analyzed statistically, as for example, to determine the values of standard, mean, maximum and minimum deviation, and autocorrelation values for any previous time interval. The newly calculated temporal parameters and the radio link parameters are then correlated with the aim of providing a set of correlated parameters more closely related to the quality of the voice. An estimator that uses the correlated parameters then calculates an estimate of the quality of the voice. In one aspect of the apparatus of the present invention, a functional apparatus for measuring the quality of the voice in a cellular telecommunication network is described. The apparatus consists of three functional steps where the first stage, a temporary processing stage, is arranged to receive a set of radio link parameters contained in a data frame transmitted from a mobile station. The temporal processor calculates a set of temporary parameters to be recorded in the correlation processing step. Correlation processing correlates temporal parameters to derive relationships between the parameters most closely related to voice quality. The correlated parameters are then recorded in an estimator stage in order to calculate an estimate of the speech quality. The estimator can be based on a linear or non-linear estimate. In addition, the estimator may consist of a neural network, or a state machine configured to change state in response to a change in a dynamic variable, such as the speed of a moving mobile station or a change of jump from frequency to absence of frequency jump. The present invention employing radio link parameters offers a simple and reliable method for measuring the quality of voice in a cellular network. In addition, the dynamic nature of the technique allows the operator to constantly update itself regarding quality conditions in all parts of the network. These and other advantages of the present invention will be apparent upon reading the following detailed description and upon studying the various figures of the drawings. BRIEF DESCRIPTION OF THE DRAWINGS The invention, together with further objectives and advantages thereof, will be better understood with reference to the following description in combination with the accompanying drawings in which: Figure 1 shows a prior art method for measuring of voice quality using signal databases; Figure 2 shows a method for measuring speech quality in a mobile communication network in accordance with an embodiment of the present invention; Figure 3 shows a block diagram of the method of measuring the quality in accordance with an embodiment of the present invention; and Figure 4 shows a graph of an exemplary parameter correlated with voice quality. DETAILED DESCRIPTION OF THE PREFERRED MODALITIES In the previous sections, a commentary of Figure 1 was offered focusing on a prior art method of measuring speech quality. In a basic cellular system, a mobile switching center (MSC) is linked to a plurality of geographically dispersed base stations (BS) to form the cellular coverage area for the system. Each of the base stations is designed to cover a specific area known as a cell, wherein a two-way radio communication can be established between a mobile station MS and the base station in the associated cell. The level of quality of coverage is not uniform for all points in the coverage area due to several uncontrollable factors. Therefore, the quality perceived by the end user offers important information regarding the current level of network performance. Below is a description of a method for measuring voice quality in the network by monitoring radio link parameters. Figure 2 illustrates the basic concept of the use of radio link parameters that are available, for example, in a mobile station, base station and mobile switching center in the case of a typical TDMA-based network. By way of example, a transmitter 12 of an associated base station transmits a signal from an antenna 14 through the air in the form of digitally modulated information packets (digital packets). In an ideal situation, the transmitted signal would be received in its original form without errors by the receiver 16 in the mobile station. In practice, a distortion caused by weak signals (shadowing), multipath fading, as well as co-channel interference may introduce transmission errors. In systems operating in accordance with D-AMPS, for example, voice or other data are sent in digital packets of 20 milliseconds, known as frames, further divided into 6 time segments. In a downlink situation, coded voice data is transmitted to a mobile station by employing two time segments in each frame and decoded in a voice decoder in the mobile station. As the frames are transmitted, bit errors introduced by distortions in the bit stream are received and said bit errors are detected by the mobile station and a bit error ratio (BER) is calculated 18. A frame containing the data can be marked as "bad" when the number of errors in the bits is greater than a specific limit or when checksum errors are detected. The rate of occurrence of "bad" designates a frame erasure rate (FER) that is reported as parameter 20. A "bad" frame, containing necessary control information and data, is not reliable, and therefore "can not" In this situation, data from a previous "good" frame are used in an attempt to recover the errors in the bits Another parameter reported by the mobile station is the received signal level (RxLev) 22 that reports the force A transfer parameter 24, which represents statistical aspects of transfer events, is reported and indicates that the call has been switched to another frequency, for example, during a transfer situation between cells to another cell. parameters 26 which contain, for example, soft information are obtained from the receiver 16. A soft information may contain, for example, information as to the quality of the bits in a A method for the use of soft information and to improve the quality estimate is presented in US Pat. No. 5,432,778 issued to Minde et al, entitled: Meted and An Arrangement for Frame Detection Quality Estimation in the Receiver of a Radio Communication System (Method and arrangement for quality estimation of frame detection in the receiver of a radio communication system) granted on 11/7/95 that is incorporated herein by reference in its entirety. An estimate of the speech quality 28 can then be made from the parameters measured in accordance with what is described below. In a cellular network, there is a predefined voice communication link, therefore, a known voice coder / decoder (codec) is used which conforms to a specified standard. The final perceived quality of the voice is affected not only by the number of errors in the bits but also by their distribution over time. For example, a noticeable fade drop may cause short packets of errors in the bit stream, which are in close temporal proximity, and may in turn cause the channel decoder to fail while decoding. This can introduce a frame erasure or it can cause erroneous voice decoding. Frame erasures can be hidden by repeating the parameter data bits of previous frames, which can result in a "synthetic" sound due to regeneration. In addition, erroneous speech decoding and synthesis due to decoding failures can propagate over some frames and can result in undesirable loud sounds. Thus, a short packet of errors in the sequential bits can cause a significant decrease in quality for a certain time. On the other hand, many fast fade drops can introduce a low average residual BER and result in a better perceived voice quality since channel decoding can correct most errors. Therefore, the foregoing suggests that the temporal characteristics of the parameters related to voice quality should be taken into account. These parameters carry information regarding different properties, for example, fading rates, fading durations, fade depths, signal-to-noise ratio, signal-to-interference ratios, signal levels as well as transfer situations. Therefore, it is possible to extract additional information regarding perceived quality from correlations and cross correlations of these parameters over time. Referring now to Figure 3, a voice quality measurement method employing temporal and correlation processing is presented, in accordance with one embodiment of the present invention. The preferred embodiment comprises a multi-stage configuration including a temporary processing stage 32, a correlation processing stage 34, and an estimator stage 36. Radio link parameters such as BER, FER, RxLev, HO, as well as as soft information they are entered in the temporary processing stage 32. From these parameters, new parameters can be calculated. As the experts in the field can observe, the temporal processing of the parameters can be carried out, for example, through the application of what is known as "window sliding" or simply "window selection" in the domain temporary as for example rectangular, exponential, and hamming (window sin2) in order to achieve a temporary weighting. The parameters can be correlated, then, taking for example the root, exponential, or logarithm of the function in order to achieve a more appropriate form. In addition, the transformed data can be analyzed with statistical methods that can include the determination of the maximum value, minimum value, average value, standard deviation, asymmetry, kurtosis, etc. These processes can be performed independently and in any order to achieve the desired relationships. A temporary block processing 32 is desirable to extract temporary information from the parameters by examining its previous activity during a specified time interval. By way of example, the examination of a history of sequence of measurements for a parameter, it is possible to calculate temporary parameters such as for example average value for the last X seconds, estimate the standard deviation during Y seconds, or the autocorrelation function during the last Z seconds. In one example, the average BER during the last 3 seconds or the number of frames erased during the last 5 seconds are parameters representing new temporal parameters to knock down parameters more closely related to a voice quality aspect. The correlation stage in block 34 correlates the original or newly calculated jsemporal parameters to produce correlated parameters that are more directly related to the quality of the voice. For example, modern cellular systems try to hide the loss of a frame due to errors in the bits by repeating the previous frame of 20 ms hoping that it will not be heard. This means that the number of errors in the bits in the lost box is not relevant since the content of the frame never reaches the person who is listening. This suggests that a new parameter that correlates more closely with voice quality can be calculated by correlating the BER with the Frame Loss, for example. In the first example, it works well with the present invention and illustrates the use of temporal processing and correlation, the mean for the BER is calculated in 0.5 second intervals, in the temporal processing stage 32 in order to create a new parameter temporary RXQ_MEAN_5. In the correlation stage 34, the RXQ_MEAN_5 parameter is correlated by the application of a third power transformation providing a correlated parameter (RXQ_MEAN_5) 3. In a second example, the FER is calculated in 0.5 second intervals in order to form the temporary parameter FER_MEAN_5. A cube root transformation is then applied to the temporary parameter FER_MEAN_5 to form a correlated parameter (FER_MEAN_5) 1/3. In a third example, the FER is calculated in a 5 second interval to determine the number of consecutive frame erasures to form the FER_BURSTS_5 parameter. A subsequent correlation is performed by applying a square root transformation to the temporal parameter in order to form a correlated parameter (FER BURSTS 5) 1/2 A summary of the associated temporal parameters and associated parameters is given below in the table A. TABLE A TEMPORARY PARAMETER CORRELATED PARAMETER RXQ_MÉAN_5 (RXQ_MEAN_5) 3 FER_MEAN_5 (FER_MEAN_5) 1/3 FER BURSTS 5 'FER BURSTS 5) 1/2 Other potential parameters may include performing operations similar to the proportion of errors in the residual bits (RBER, where RBER is equal to zero when the frame is cleared, and is equal to the BER when the frame is not cleared) and other parameters received. It will be noted that temporal processing and statistical analysis can be performed on the correlated parameters and that some, for example, RBER can be calculated in "raw" data. An estimator stage in block 36 employs the correlated parameters to calculate an estimate of the perceived speech quality. The estimator 36 can be based on several mathematical models, such as linear, non-linear, or they can comprise a neural network. A simple linear model can have the form: Estimation = A (parameter 1) + B (parameter 1) + .... where the coefficients A and B are optimized to obtain the best performance. The coefficients may be derived, for example, by the use of a linear regression technique on a subjectively graded preparation material. Even when a linear estimate offers adequate results, as can be observed by a person skilled in the art, nonlinear estimators can offer a more accurate estimate. An exemplary procedure employing a linear estimate can be carried out on the correlated parameters of an example above and can have the form of: Estimation = A * (FER_MEAN_5) 1 3 + B * where the coefficients A and B can be derived by techniques of linear regression mentioned above that are well known. In addition, it is possible to combine any number and combination of radio link, temporal or correlated parameters for the estimation of conformity with that determined as optimal for several situations by a person skilled in the art. In addition, specific examples of temporal and correlated parameters have been provided and accordingly various modifications to the described parameters can occur to those skilled in the art and said variations are within the spirit and scope of the present invention. Particularly, modifications in relation to temporal and correlated parameters as well as variations in interval durations can be changed to suit the particular type of interference or situation experienced. A non-linear estimate can also be carried out by several linear estimators that approximate the almost linear portions of a modeled curve. Figure 4 represents a graph of the relationship between Quality (Q) and carrier-to-interference ratio (C / I) using this technique. The curve 60 can be divided into several almost linear segments to be modeled with the successive linear estimators. For example, segment 62 is strongly inclined and has little curve, and therefore can be represented by a linear model. Similarly, segment 64 has a slightly more pronounced curve and can also be approximated through a linear model. Segment 66 of the curve begins to flatten and can be approached very adequately with a linear model. In order to offer a seamless transition between the models, it is necessary to determine where the current operating point is located. One method that can be used to solve this is the use of a model to determine the probability of being in a specific segment. The linear models used in the multiple estimator approach can offer relatively simple and accurate modeling.

In addition, a multi-stage neural network can be used which produces more accurate results. Neural networks are networks of processors or neurons joined by unidirectional connections that carry data and are weighted accordingly. Neurons act independently and operate based only on their inputs by associated weighting. Typically, neural networks require preparation algorithms to adjust the weights based on the presented patterns. For example, a preparation technique that can be applied to a neural network estimator is by simultaneous recording of radio link parameters with test speech. Registered voice is evaluated by a panel of people who are listening where they qualify. By way of example, the radio link parameters are processed in the temporal processing stage 32 and in the correlation processing stage 34 of FIG. 3 where the result plus the qualifications for preparing the network are used. As is known to those skilled in the art, one advantage of using a neural network is that processing in steps 32 and 34 may be less complicated since the network may be better suited for this task than ordinary estimators. An example of a neural network that can be employed with the present invention is offered in U.S. Patent No. 5,432,778, the disclosure of which is incorporated herein by reference.

In addition, according to the characteristics of the system, said carrier frequency and frequency hopping, another type of estimator that may be suitable is an estimator based on a finite state machine that changes state in accordance with some dynamic criteria. For example, the estimator can be configured to change state in response to a change in the mobile speed or a change from the frequency hop to no frequency hop and vice versa. By way of example, this may be appropriate in situations in which the model may be different, for example, in the case of a call with frequency hopping compared to a call without frequency hopping. The present invention contemplates a method for measuring the quality of the voice in a cellular telecommunication system by monitoring the radio link parameters. The above discussion also encompasses an inherently simple and accurate voice quality estimation technique that avoids the complexities associated with voice databases and perceptual models. The present invention exploits the use of the temporal information of the current radio link parameters by calculating new parameters in which relationships and cross-correlations between parameters can be used in order to obtain an improved estimate of the voice quality. Although the invention has been described in some aspects with reference to a specific preferred embodiment, various modifications and applications thereof will be apparent to those skilled in the art. Particularly, the concept of the present invention can be applied, in addition to D-AMPS, to other digitally based systems operating in accordance with, for example, Code Division Multiple Access (CDMA), Global System for Mobile Communication (GSM), or Personal Digital Cell (PDC). Therefore the intention is that the following claims do not receive a restrictive interpretation but should be considered as encompassing variations and modifications derived from the subject matter of the presented invention.

Claims (21)

1. CLAIMS A method for estimating voice quality _in a radio telecommunication system comprising the steps of: receiving a set of radio link parameters; processing the temporal information of said radio link parameters in order to calculate a set of temporal parameters; correlating said parameters to produce a set of correlated parameters; and estimate the voice quality from the parameters correlated with an estimator. A method according to claim 1 wherein the radio link parameters include at least one of the following ": BER, FER, RxLev, transfer statistics, soft information, as well as speech energy parameters. Claim 1 wherein the step of extracting temporary information further comprises the step of transforming the parameters into the time domain by performing any of the following: power, exponential, logarithmic, hamming window, exponential, or rectangular window operations. A method according to claim 2 wherein the step of temporary processing includes the calculation of the average BER in a 0.5 second interval and where said correlation step includes the application of a cubic transformation to said average BER. claim 2 wherein the temporary processing step includes the calculation of the Mean FER in a 0.5 second interval and where said correlation step includes the application of a cubic root transformation to said average FER. A method according to claim 2 wherein the step of temporary processing includes the calculation of the Average FER in an interval of 5 seconds and where said correlation step includes the application of a square root transformation to said average FER. A method according to claim 1 wherein a time interval duration within a range of about 0.1 to 10 seconds is employed in said temporary processing step. A method according to claim 1 wherein the estimation step is carried out through a linear estimator. A method according to claim 1 wherein the estimation step is carried out through a non-linear estimator. A method according to claim 1 wherein the estimation step is carried out through a neural network. 11. A method according to claim 1 wherein the estimation step is carried out through multiple linear estimators. 12. A method according to claim 2 wherein the estimation step is carried out through a state machine estimator, wherein the estimator is configured to change state in response to a change in any of said parameters. 13. A method according to claim 12 wherein the estimation is carried out through a state machine estimator, where the estimator is configured to change state in response to a change in the moving speed, 14. A method according to claim 13 wherein the estimator is configured to change state in response to a change of frequency hopping to absence of frequency hopping and vice versa. 15. A system for measuring voice quality in a radio telecommunications network, comprising: a temporary processor for extracting temporal information from a set of available radio link parameters associated with a received radio signal, wherein said Temporal processing generates a set of temporal parameters that are related to the quality of the voice; a correlation processor to determine the correlation between suitable radio link parameters and temporal parameters in order to generate a set of correlated parameters; and an estimator for determining an estimate of voice quality from suitable generated parameters and radio link parameters. 16. An apparatus according to claim 15, where the radio link parameters include at least one of the following: BER, FER, RxLev, transfer statistics, soft information, and voice energy parameter. 17. An apparatus according to claim 15, wherein the estimator consists of a linear estimator. 18. An apparatus according to claim 15, wherein the estimator consists of a non-linear estimator. 19. An apparatus according to claim 15, wherein the estimator consists of a neural network. 20. An apparatus according to claim 15, wherein the estimator consists of a state machine configured to change state in response to a variable event such as, for example, velocity of a mobile station in motion. 21. An apparatus according to claim 20 wherein the state machine is configured to change state in response to a change of frequency hopping to absence of frequency hopping and vice versa. An apparatus according to claim 15 wherein the estimator consists of multiple linear estimators. SUMMARY OF THE INVENTION A method and system for measuring voice quality in a mobile cellular telecommunication network by using available radio link parameters is presented. In a preferred embodiment, the method includes receiving a set of radio link parameters, in accordance with what is defined in a standard or otherwise available, such as BER, FER, RxLev, transfer statistics, soft information , and voice energy. A temporal information is obtained from radio link parameters in order to create a set of temporal parameters that can be analyzed statistically, for example, for the values of standard deviation, mean, maximum and minimum, and values of autocorrelation by a time interval. The temporal parameters are combined in order to provide a set of correlated parameters more closely related to the quality of the voice. An estimator then uses the correlated parameters in order to calculate an estimate of the quality of the voice. The method of the present invention takes advantage of the temporal information and the correlated relationships of the transmitted parameters. In addition, the method is inherently simple and reliable in comparison with the methods of the prior art. In addition, the technique. It offers a method and allows an operator to monitor quality conditions in the network.