KR100347752B1 - Apparatus and Method for objective speech quality measure in a Mobile Communication System - Google Patents

Abstract

The present invention defines the Bark Coherence Function by reflecting the human auditory characteristics, and by using this objective objective sound quality evaluation device in the mobile communication system to enable the objective sound quality evaluation having a high correlation with the subjective sound quality And a method thereof, wherein the present invention calculates a bark spectrum of a voice signal input into a cellular system, calculates a bark spectrum of a voice signal output from the cellular system, and mutual barks of the input and output voice signals. Calculate the spectrum. Then, Bark coherence function is calculated by substituting the calculated Bark and Cross Bark spectra into the Bark coherence function, and the calculated result is converted into the Bark noise-to-signal ratio. .

Description

Apparatus and Method for objective speech quality measure in a Mobile Communication System}

TECHNICAL FIELD The present invention relates to an objective speech quality measure in a mobile communication system. In particular, the present invention defines a Bark Coherence Function reflecting human auditory characteristics and correlates subjective sound quality with the subject. The present invention relates to an objective sound quality evaluation apparatus and a method in a mobile communication system capable of highly objective sound quality evaluation.

In line with the human desire to communicate freely regardless of time and place, mobile phones have become widespread in recent years. The most important factors for determining the quality of such a mobile phone are the call success rate and the call quality, and in particular, the call quality, that is, the sound quality, is determined by subjectivity according to the user's hearing. Therefore, continuous sound quality evaluation is essential for the installation, maintenance and repair of mobile telephone networks.

Currently, the voice signals transmitted through the mobile telephone network are represented as digital signals compressed by the voice coder, and the importance of these signals in the sound quality of the reproduced voice signals varies greatly. Therefore, the sound quality of mobile telephone network should be evaluated by considering the characteristics of voice signal.

The sound quality evaluation may be basically performed by a subjective sound quality evaluation method through repeated listening experiments of various users. However, this method is not suitable because it takes a lot of time, effort, and cost to perform it repeatedly in various environments despite the advantage that it is directly related to the user's haptic sound quality. Therefore, it is desirable to predict the subjective sound quality on an objective scale having a high correlation with the subjective sound quality.

On the other hand, sound quality evaluation includes a subjective sound quality evaluation method in which a sound quality evaluation experimenter directly listens to a sound signal to be evaluated and evaluates sound quality according to subjective judgment, and an objective sound quality evaluation method using an algebraic difference between the original sound signal and the distorted sound signal by calculation. have.

First, the subjective sound quality evaluation method is roughly classified into an interactive evaluation method by two-way listening and a listening method by one-way listening. The dual interactive evaluation method can be performed only when the complete communication system is completed, and cannot be applied during the development stage of the speech coder or the installation stage of the mobile telephone network. Therefore, audible evaluation that is easy to perform is commonly used.

Listening evaluation methods include Absolute Category Rating (ACR), which listens only to distorted speech signals and evaluates them, Degradation Category Rating (DCR), which listens to the original and distorted speech signals and evaluates the degree of distortion. There is a comparison category rating (CCR) that listens and compares two distorted speech signals.

Among them, the MOS (Mean Opinion Score) evaluation method is used for the absolute sound quality evaluation method which evaluates the sound quality without the comparison voice signal most similarly to the actual telephone usage environment. The MOS evaluation method is a method in which the experimenter hears the distorted speech signal and evaluates the distortion degree of the speech signal in five steps as shown in [Table 1] below.

MOS rating Rate Speech Quality Level of Distortion 5 Excellent Imperceptible 4 Good Just perceptible but not annoying 3 Fair Perceptible and slightly 2 Poor Annoying but not objetionable One Bad Very annoying and objectionable

However, as the subjective sound quality evaluation method is well-known, it is directly assessed by a person, and thus has the advantage of expressing the actual sound quality of a real user best, but it has a disadvantage of requiring time and cost.

In order to improve the shortcomings of the subjective sound quality evaluation method, the sound quality evaluation method that has been in the spotlight recently is an objective sound quality evaluation method.

The objective sound quality evaluation method can be classified into an evaluation in the time domain, an evaluation in the frequency domain, and an evaluation in the psychoacoustic domain according to the evaluation region.

First, objective sound quality evaluation in the time domain includes signal-to-noise ratio (SNR) and segment-to-noise ratio (SegSNR). These methods use the squared mean error between the original speech signal and the distorted speech signal, and are suitable for the performance evaluation of waveform encoders having a bit rate of 32 Kbps or more.

Objective sound quality evaluation in the frequency domain includes spectral distance (SD) and algebraic spectral distance (LSD). Among these methods, LPC-CD calculates the difference between the envelope components of speech signals and is suitable for the performance evaluation of speech coders with 16 to 32 Kbps data rates.

In the psychoacoustic field, objective sound quality evaluation is a method of measuring distortion of two speech signals by using a psychoacoustic model that reflects the human auditory characteristics of the original speech signal and the distorted speech signal. BSD (Bark Spectral Distortion), Perceptual Linear Predictive-Cepstum Distance (PLP-CD), Measuring Normalizing Blocks (NBN), Perceptual Speech Quality Measure (PSQM), and the like. In particular, PSQM and MNB are recommended by ITU-T, and PSQM is suitable for performance evaluation of speech coders of 4Kbps or more, and MNB is suitable for performance evaluation of speech coders with bit errors and frame deletions.

PSQM, MNB of the objective sound quality evaluation method in the psychoacoustic domain in more detail, ITU-T (International Telecommunication Union-Teeelecommunication) In PSCM in Rec.P.861 as an objective sound quality evaluation method, Rec.P .861 Appendix II recommends MNB. The PSQM is suitable for the objective sound quality evaluation of the CELPCode Excited Linear Prediction (CLP) speech coder of 4 kbps or more, and the MNB is suitable for frame deletion of the speech coder speech packet and cell loss of Asynchronous Transfer Mode (ATM). PSQM and MNB transform the original and distortion spectrums into the psychoacoustic region by using the human auditory characteristics.

Both methods have been developed for speech coders, so performance is degraded when evaluating the sound quality of a real mobile phone or packet voice transmission system. The actual voice delivery system consists of an analog system in addition to a digital system. Analog systems, such as landline telephones, cause linear distortion, which has little effect on subjective sound quality, but affects the objective sound quality and degrades the performance of the objective sound quality evaluation system.

1 is a diagram illustrating a distribution of MOS and PSQM of speech that has passed through a CDMA PCS system. FIG. 1A is a MOS distribution diagram, and FIG. 1B is a PSQM distribution diagram.

Each voice data is divided into two groups, Group A and Group B. These two crowds, Group A, are composed of voices with a slightly reduced low frequency range through a private exchange. The two groups have similar MOS distributions, indicating that voices passing through analog systems with different passbands have similar sound quality.

However, the PSQM, an objective sound quality evaluation measure known to have a high correlation with subjective sound quality, obtains a difference in spectrum, and thus has a different spectral distance in the low frequency region, resulting in a different distribution. If you evaluate the data in group B by regression analysis with group A data, you will not get a consistent result. For example, in group A, if the PSQM value is 4.5 and relatively low in voice quality, the subjective sound quality is predicted by the regression function trained in group B. In group B, the value of 4.5 belongs to the region with poor sound quality. As a result, the prediction is wrong.

Therefore, the present invention is proposed to solve all the problems that occur in the conventional objective sound quality evaluation as described above,

An object of the present invention is to define the Bark Coherence Function by reflecting the human auditory characteristics, and to use it to objectively evaluate the objective sound quality having a high correlation with the subjective sound quality. The present invention provides a sound quality evaluation device and method thereof.

The present invention for achieving the above object,

To overcome the effects of linear distortion in analog systems, we define a Bark Coherence Function (BCF), calculate the Bark spectrum and the cross-bark spectrum for each of the input and output signals of the cellular system. The Bark coherence function is obtained by applying the Bark spectrum and the mutual Bark spectrum to the Bark coherence function. Then, the resultant value is converted to the Bark signal-to-noise ratio to calculate the sound quality evaluation value, thereby overcoming the influence of linear distortion, and objective sound quality evaluation having a high correlation with the subjective sound quality is possible.

1 is a diagram illustrating the distribution of MOS and PSQM of voice that has passed through a conventional CDMA PCS system, FIG. 1A is a distribution diagram of MOS, and FIG. 1B is a distribution diagram of PSQM.

2 is a block diagram showing an objective sound quality evaluation apparatus in a mobile communication system according to the present invention,

3 is a flowchart illustrating an objective sound quality evaluation method in a mobile communication system according to the present invention;

4 is a flowchart illustrating a bark spectrum calculation process of the audio signal of FIG. 3;

5 is a flowchart illustrating a process of calculating a cross bark spectrum of the voice signal of FIG. 3;

6 is a characteristic diagram of a psychological weighting filter applied to the present invention,

7 is a view comparing the BCF distribution according to the present invention and the conventional MOS distribution, Figure 7a is a conventional MOS distribution, Figure 7b is a BCF distribution according to the present invention,

8 is a diagram comparing the correlation coefficient of the PSQM applied to the conventional objective sound quality evaluation method and the correlation coefficient of the BCF according to the present invention.

<Explanation of symbols for the main parts of the drawings>

10: cellular system

20, 30: first and second bark spectrum calculation unit

40: mutual bark spectrum calculation unit

50: Bark coherence function calculation unit

60: Bark noise to signal ratio calculation unit

Hereinafter, a preferred embodiment of the present invention according to the technical spirit as described above will be described in detail with reference to the accompanying drawings.

In general, Magnitude-Squared Coherence (MSC) represents a correlation between two signals in the frequency domain and is defined as in Equation 1 below.

Where S _xy (f) is the cross spectrum between the input speech signal x (n) and the output speech signal y (n), and S _xx (f) and S _yy (f) are the input speech signal x (n) and output speech, respectively. Auto spectrum of the signal y (n). In an ideal environment without noise Has a value of 1, on the contrary, if there is only noise, the MSC has a value of zero. These MSCs are used in various fields such as objective sound quality evaluation, system identification and response, propagation path estimation, and delay measurement.

Psychoacoustics allows you to use MSC in the sound domain where the human feels directly in the ear, unlike the strength of physical sounds. BCF, an MSC defined in the psychoacoustic domain, is characterized by the critical psycho-acoustic characteristics of critical-band spectral resolution, equal-loudness preemphasis, and intensity-loudness power law. Use

That is, the critical band spectral resolution reflects the masking characteristic of the tone masked by noise. The threshold band filter consists of several filters with a passband width that increases on a logarithmic scale with a nonlinear Bark frequency. The human ear has different sensitivity at each frequency. In other words, human frequency resolution is more sensitive to low frequency region than high frequency region. For example, a 100 Hz tone must have 35 dB greater intensity to sound the same as a 1000 Hz tone. This property is already well known as the equal curve and using this is the equal curve correction. The intensity of the sound after the equalization curve is expressed as a phone, which means that the sound is corrected to be the same size as the tone of 1 kHz. Subjective sound correction reflects the nonlinearity of auditory intensity. This indicates that the loudness needed to feel doubled is not always constant at the phone level, but depends on the loudness level. The pawn unit is converted into a unit of son, which represents the actual subjective intensity level. The Bark spectrum is the transformation of the spectrum of speech using these three characteristics.

The present invention defines the general Bark spectral function as described above and extracts the sound quality evaluation value for the objective sound quality evaluation.

Example 1

2 is a block diagram showing the configuration of an objective sound quality evaluation apparatus according to the present invention.

Here, reference numeral 10 denotes an experimental cellular system, reference numeral 20 denotes a first bark spectrum calculator for obtaining a bark spectrum of an audio signal input to the cellular system 10, and reference numeral 30 denotes the cellular system 10. Represents a second Bark spectrum calculation unit for calculating the Bark spectrum of the audio signal output from the reference signal, and reference numeral 40 denotes a cross bark spectrum of the voice signal input to the cellular system 10 and the voice signal output from the cellular system 10. A cross bark spectrum calculation unit to be obtained is represented, and reference numeral 50 denotes each bark spectrum obtained from the first and second bark spectrum calculation units 20 and 30 and the cross bark spectrum obtained from the cross bark spectrum calculation unit 40. Bark Coher to compute the Bark Coherence function by assigning it to the Bark Coherence function A reference function calculation unit 60 denotes a bark noise-to-signal ratio calculation unit for converting the signal output from the bark coherence function calculation unit 50 into a bark noise-to-signal ratio and outputting the result as an objective sound quality evaluation value. .

In the mobile communication system according to the present invention configured as described above, the objective sound quality evaluation device first outputs the cellular system 10 after processing the input voice signal x (n) in the same manner as a general voice processing method. . The voice signal y (n) output here is a linearly distorted voice signal.

Next, the first bark spectrum calculator 20 calculates a bark spectrum of the voice signal x (n) input to the cellular system 10. That is, a window function is put on the input signal, fast Fourier transform (FFT) of the signal in the window function, and then the power power spectrum of the voice is obtained. The obtained power spectrum is then filtered using a psychometric weighting filter. Here, the psychological weighting filtering is obtained by the weighted sum of the psychological weighting filters in the frequency domain as shown in FIG. The psychometric weighting filter takes into account the characteristics of critical band spectral resolution and equality curve correction among human psychoacoustic characteristics. The subjective loudness of the signal thus obtained is corrected and then the result is output as the bark spectrum L _xx (i) for the input speech signal. The Bark spectrum, which is the result of the calculation, refers to the spectrum of speech felt in the human auditory region.

In addition, the second bark spectrum calculator 30 calculates a bark spectrum of the voice signal y (n) output from the cellular system 10. That is, a window function is put on the output speech signal, fast Fourier transform (FFT) of the signal in the window function, and then the power spectrum of the speech is obtained. The obtained power spectrum is then filtered using a psychometric weighting filter. The subjective loudness of the signal thus obtained is corrected and then the result is output as the bark spectrum L _yy (i) for the output speech signal.

In addition, the mutual bark spectrum calculator 40 calculates the mutual bark spectrum of the input voice signal x (n) and the output voice signal y (n). That is, a window function is applied to the two input signals x (n) and y (n), respectively, and the signals in the window function are fast Fourier transformed, respectively. Then the cross power spectrum of the two speech signals is obtained. The obtained signal is then filtered with a psychometric weighting filter, and then the subjective loudness of the filtered signal is corrected and the result is output as a cross-bark spectrum (L _xy (i)).

Accordingly, the bark coherence function calculation unit 50 may include the bark spectra L _xx (i) and L _yy (i) output from the first and second bark spectrum calculation units 20 and 30, respectively. The bark coherence function is substituted by substituting the cross bark spectrum L _xy (i) output from the cross bark spectrum calculation unit 40 into the following equation (2), that is, the bark coherence function BCF (i). Will be obtained.

Where i represents the Bark frequency.

Next, the bark noise-to-signal ratio calculation unit 60 calculates a bark noise to signal ratio (BDSR) using the result obtained from the bark coherence function calculating unit 50, and the result value. Is output as an objective sound quality evaluation value.

Here, BDSR represents the distortion degree of the distorted sound with respect to the original sound in psychoacoustic sound, as shown in Equation 3 below.

7 is a view comparing the BCF distribution according to the present invention and the conventional MOS distribution, FIG. 7A is a conventional MOS distribution, and FIG. 7B is a BCF distribution according to the present invention. Unlike FIG. 1, it can be seen that the distributions of BCF values of group A and group B that pass through the PSTN channel having different characteristics are similar to each other. In other words, the BCF uses the MSC in the psychoacoustic domain, so unlike the conventional objective sound quality evaluation using the average Euclidean distance, the BCF is not affected by the analog system which does not affect the subjective sound quality.

8 is a diagram comparing the correlation coefficient of the PSQM applied to the conventional objective sound quality evaluation method and the correlation coefficient of the BCF according to the present invention.

In order to evaluate the performance of the proposed BCF, the speech data and the regression analysis used in FIG. The performance of the objective sound quality evaluation is evaluated by the Pearson's correlation coefficient between the actual subjective sound quality evaluation result and the predicted subjective sound quality.

PSQM, an ITU-T standard, has a high correlation coefficient for group A and B separately, but a very low correlation coefficient for group A and B. Since BCF is not affected by linear distortion, it has a high correlation coefficient in all cases. Therefore, it can be seen that BCF is suitable for evaluating the sound quality of the actual data passing through the cellular system.

Example 2

3 is a flowchart illustrating an objective sound quality evaluation method in a mobile communication system according to the present invention.

As shown therein, calculating the bark spectrum of the input voice signal (S10), calculating the bark spectrum of the output voice signal (S20), and calculating the mutual bark spectrum of the input and output voice signals. (S30) and calculating the Bark coherence function by substituting the calculated two Bark spectra and the mutual Bark spectra into the Bark coherence function (S40), and the resultant values obtained in the step S40 vs. Bark noise And outputting the result value obtained by converting the signal ratio as an objective sound quality evaluation value (S50).

The objective sound quality evaluation method thus obtained first calculates the bark spectrum of the voice signal input to the cellular system in step S10. The bark spectrum calculation of the input voice signal is performed by applying a window function to the input voice signal in step S11 as shown in FIG. 4, and performing fast Fourier transform (FFT) on the signal in the window function in step S12. Obtain In step S13, the power spectrum of the voice is psychometric weighted and converted into an objective luminance region, and in step S14, the bark spectrum of the final input voice signal is calculated through the luminance correction.

Next, a bark spectrum of the distorted output speech signal output from the cellular system is calculated in step S12. The calculation of the bark spectrum of the audio signal output here calculates the bark spectrum of the output audio signal in the same manner as the calculation of the bark spectrum of the input audio signal shown in FIG. 4. That is, a window function is put on the output voice signal, and the signal in the window function is fast Fourier transformed (FFT) to obtain a power spectrum of the voice. The power spectrum of the obtained speech is psychometric weighted and converted into an objective luminance region, and the bark spectrum of the final output speech signal is calculated through the luminance correction.

Next, in step S30, the mutual bark spectrum of the input speech signal and the output speech signal is calculated. That is, as shown in FIG. 5, the cross-bark spectrum calculation obtains a cross power spectrum of speech after fast Fourier transforming the signals in the window function in step S31, respectively, and in step S32. In step S33, psychometric weighting is performed to convert to the objective luminance region, and in step S34, the mutual bark spectrum of the final input and output voice signals is calculated through subjective sound intensity correction.

Next, the Bark coherence function is calculated by substituting the two Bark spectra and the cross Bark spectrum calculated in step S40 into the Bark coherence function of Equation 2, which is obtained. The resulting value is converted to the Bark noise-to-signal ratio to output a sound quality estimate.

According to the above-described "objective sound quality evaluation apparatus and method in the mobile communication system", the objective sound quality evaluation having a high correlation with the subjective sound quality regardless of the analog environment is possible, and the system construction and customer satisfaction When applied to the part that needs continuous sound quality evaluation to provide a service for not only accurate sound quality evaluation, but also has the advantage of reducing the time and cost required for sound quality evaluation.

Claims (12)

delete delete delete delete delete delete In the objective sound quality evaluation method in a mobile communication system, Applying a window function to an input speech signal, performing fast Fourier transform (FFT) on the signal in the window function, obtaining a power spectrum of the speech, and psychometric weighting the power spectrum of the speech to an objective luminance region A first step of calculating a bark spectrum of a final input speech signal by performing a luminance correction of the signal on which the psychological weight filtering has been performed; Applying a window function to an output speech signal, performing fast Fourier transform (FFT) on the signal in the window function, obtaining a power spectrum of the speech, and psychometric weighting the power spectrum of the speech to an objective luminance region A second step of calculating a bark spectrum of a final output speech signal by performing a luminance correction of the signal on which the psychometric weighting filtering has been performed; Applying a window function to the input / output speech signal, obtaining a cross power spectrum of the speech after fast Fourier transforming the signal within the window function, and psychometric weighting filtering the cross power spectrum of the speech A third step of calculating a cross-bark spectrum of a final input and output speech signal by performing a step of converting to and performing a luminance correction on the signal on which the psycho-weighted filtering is performed; A fourth step of obtaining the Bark coherence function by substituting the calculated Bark and mutual Bark spectra into the Bark coherence function; And a fifth process of converting the resultant value obtained in the fourth process into a bark noise-to-signal ratio and outputting the resultant value as an objective sound quality evaluation value. delete delete delete The Bark coherence function of claim 4, wherein the Bark coherence function calculation of the fourth process comprises: a Bark coherence function of each of the Bark spectra obtained in the first and second processes and the cross Bark spectrum obtained in the third process. An objective sound quality estimation method in a mobile communication system, comprising calculating a Bark coherence function by substituting for. Where L _xy (i) is the cross-bark spectrum, L _xx (i) is the bark spectrum of the input voice signal, L _yy (i) is the bark spectrum of the output voice signal, and i is the bark frequency. The method of claim 7, wherein the Bark noise-to-signal ratio calculation of the fifth process comprises converting a signal obtained in the fourth process into a Bark noise-to-signal ratio calculation equation as follows and outputting the result as an objective sound quality evaluation value. Objective sound quality evaluation method in a mobile communication system, characterized in that. Where i represents the Bark frequency and BCF represents the value calculated by the Bark coherence function.