JP2005136803A - Fault detecting apparatus for duplexed sound signal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To detect a fault in real time even when the fault occurs only in a sound signal of one system in the duplex transmission of sound signals. <P>SOLUTION: Feature quantities of small areas of the sound signals transmitted by an active system and a reserve system are extracted by signal feature quantity calculation parts 6-1, 6-2. A feature quantity comparison part 7 compares the extracted feature quantities between the systems and determines that the fault occurs in either of the systems. Feature quantity difference calculation parts 9-1, 9-2 calculate differences between the feature quantity of a fault area and the feature quantity of a normal area in each system and a [D] comparison part 10 and a fault system determination part 11 determines that the fault occurs in a system of which the calculated difference is larger. The reliability of determination is further enhanced by providing a majority processing part 12 and a significance determination part 13. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a fault detection apparatus for a duplex voice signal, and is particularly suitable for the case where the same voice signal is transmitted by being divided into two lines, and a fault-free voice signal is selectively received on the receiving side. The present invention relates to a fault detection apparatus for a duplex audio signal.

  Conventionally, in duplex transmission for the purpose of high-reliability transmission of TV signals, the same video / audio signal is divided into two lines and transmitted, and one of the lines is selected and received on the receiving side. .

  FIG. 6 is a block diagram showing a conventional duplex transmission apparatus. In the figure, only the audio signal is shown. If the upper side in the figure is the active system and the lower side is the standby system, the receiving side manual change-over switch 3 is connected so as to select the active system in the normal state. At this time, the input voice signal is transmitted through the encoder 1-1, the working transmission line 4-1, the decoder 2-1, and the manual changeover switch 3 to become an output voice signal.

  The monitoring person monitors the line status of the two systems on the receiving side, and switches the manual changeover switch 3 to the standby side when detecting that a failure has occurred in the active system. The input voice signal is transmitted through the encoder 1-2, the spare transmission path 4-2, the decoder 2-2, and the manual changeover switch 3, and the voice signal is continuously received through a normal line.

The present applicant has proposed in Japanese Patent Application Laid-Open No. 2004-228688 a technique for automating switching to a normal line when a failure occurs on a line being received in duplex transmission and shortening the failure time of the output signal as much as possible.
JP 2000-350238 A

  If the supervisor manually switches the switch and selects the line so that a normal audio signal is received, it takes time to complete the switch to the line that can receive the normal audio signal, and the output audio signal There is a problem that it is difficult to shorten the failure time.

  The technique of the above-mentioned Patent Document 1 is intended for video, and detects that there is a fault in the video and switches to a normal line. However, even if the video is normal, there is a case where a failure occurs only in the sound. In the technique disclosed in Patent Document 1, switching to a normal line is not performed in such a case, and the sound remains in the failure. There is a problem.

  An object of the present invention is to make it possible to detect a failure in real time even when a failure occurs only in one of the audio signals in the duplex transmission of the audio signal. It is possible to instantaneously switch to a normal line so that normal voice is continuously received.

  In order to solve the above-mentioned problems, the present invention provides a fault detection device for a duplex audio signal transmitted by two lines, corresponding to each system, and for each small area of the audio signal of each system. Based on the comparison result of the feature quantity comparing means, the feature quantity comparing means for comparing the feature quantities extracted by the feature quantity extracting means between the systems, and the feature quantity comparing means There is a first feature in that it includes a determination unit that determines that a failure has occurred.

  In addition, the present invention provides a fault detection apparatus for a duplicated audio signal transmitted through two lines, which is provided corresponding to each system, and extracts a feature amount for each small region of the audio signal of each system. When a failure has occurred in any of the systems and the feature region extraction means and the failure region in the duplicate audio signal is clear, the difference between the feature amount of the failure region and the feature amount of the normal region in each system is calculated. There is a second feature in that a calculation means for calculating and a fault system determination means for determining that a fault has occurred in a system having a larger difference calculated by the calculation means.

  Further, according to the present invention, the fault system determination unit uses a plurality of determination results based on the difference calculated by the calculation unit, performs a majority decision thereof, and makes a final determination. A third feature is that the determination result when the calculated difference is equal to or less than a predetermined value is made insignificant.

  In addition, the present invention further detects whether or not the sound signal in the small area is in a silence state and compares the sound signals in the two systems to detect whether or not the silence state is due to a failure. -There is a fourth feature in that a mute detection means is provided.

  In addition, the present invention has a fifth feature in that it further includes an abnormal value detecting means for detecting whether or not the feature quantity extracted by the feature quantity extracting means shows an abnormal value.

  Furthermore, the present invention is a monotonous method that preserves or inverts the sign of an input audio signal, increases the absolute value of a portion having a relatively small absolute value, and decreases the absolute value of a portion having a relatively large absolute value. A sixth feature is that a conversion means for performing conversion with an increase or monotonous decrease conversion characteristic is provided.

  According to the present invention, it is determined that a failure has occurred in one of the two systems using the feature amount of each audio signal of each system, so that a method for directly comparing audio samples can be used. As a result, stable fault detection that is less affected by coding noise or the like is possible.

  In addition, since the system in which the failure has occurred is determined based on the difference in the feature quantity for each small area of the audio signal of each system, it does not depend on the number of transmission errors compared to the method of comparing the value of the feature quantity itself. Stable determination is possible.

  In addition, using a plurality of determination results based on the difference of the feature amount for each small area of the audio signal of each system, and finally performing the majority process to determine the system in which the failure has occurred, By making the determination result based on the difference less than the predetermined value insignificant, the reliability of the determination can be improved.

  In addition, it detects whether or not the sound signal in the small area is in a silent state, and compares the sound signals in the two systems to detect whether or not the silent state is due to a failure. It is possible to prevent misidentification of other silences.

  Further, when the feature amount of the audio signal clearly shows an abnormal value, the state can be determined quickly without performing comparison operation between the feature amounts by distinguishing and detecting the feature value.

  Furthermore, the audio signal is converted with a monotonically increasing or monotonically decreasing conversion characteristic that preserves or inverts the sign, increases the absolute value of the part with a relatively small absolute value, and decreases the absolute value of the part with a relatively large absolute value. By executing the failure detection process on the converted audio signal, it is possible to accurately detect the failure even with a small audio signal.

  Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a fault detection apparatus for a duplicated audio signal according to the present invention. First, an audio signal is distributed to two paths (systems) by a distributor (not shown). When audio signals are digitally compressed and transmitted, encoders (encoders) 1-1 and 1-2 and decoders (decoders) 2-1 and 2-2 are arranged in each system.

  Here, if the upper side in the figure is the active system (system A) and the lower side is the standby system (system B), the selector switch 3 on the receiving side is connected to select the active system in the normal state. At this time, the audio signal is transmitted through the encoder 1-1, the working transmission line 4-1, the decoder 2-1, and the changeover switch 3 to become an output audio signal.

  When a failure occurs in the active system, the changeover switch 3 is switched so as to select the standby system based on the failure detection described later. This time, the audio signal is transmitted through the encoder 1-2, the spare transmission path 4-2, the decoder 2-2, and the changeover switch 3.

  In the following, detection of a failure in an audio signal will be described. First, the audio signals of the two systems A and B received on the receiving side are input to the signal logarithmic conversion processing units 5-1 and 5-2, respectively. The signal logarithmic conversion processing units 5-1 and 5-2 logarithmically convert the input audio signal while maintaining the code.

  Since the characteristics of the human ear have non-linear characteristics such that they can be discerned sensitively even if the signal level is small, it is desirable that the signal processing also conforms thereto. The signal logarithmic conversion processing units 5-1 and 5-2 have conversion characteristics satisfying this, and this conversion can change a signal change at a minute part in the audio signal to a larger change, which is an obstacle to a small sound. The detection characteristics can be improved.

  FIG. 2 shows an example of conversion characteristics of the signal logarithmic conversion processing units 5-1 and 5-2. The relationship between the input signal s and the output signal t in this conversion characteristic is expressed by the following equation (1).

  The example of FIG. 2 shows a monotonically increasing (upward to right) conversion characteristic in which the logarithmic conversion is performed while preserving the sign, but it is also possible to invert the sign and use a monotonically decreasing (downward to right) conversion characteristic. In short, with respect to the input audio signal, the sign is preserved or inverted, the absolute value of the portion with a relatively small absolute value is increased, and the absolute value of the portion with a relatively large absolute value is decreased, monotonically increasing or decreasing monotonically. Any conversion characteristic may be used.

  The audio signal logarithmically converted by the signal logarithmic conversion processing units 5-1 and 5-2 is input to the signal feature amount calculation units 6-1 and 6-2. The signal feature amount calculation units 6-1 and 6-2 calculate and extract feature amounts for each small region of the audio signal. As the feature amount, an average value or standard deviation of the audio signal can be used, and a plurality of types may be extracted.

  Next, the extracted feature quantity is input to the feature quantity comparison unit 7, and the feature quantity is compared between the two systems A and B for each small region. When the difference in the feature amount between the two systems A and B is equal to or less than a predetermined threshold value, it can be determined that no failure has occurred in any system in that region. On the other hand, when the difference in the feature amount between the two systems A and B exceeds a predetermined threshold, it can be considered that a failure has occurred in any system in that region. Based on this comparison result, it is determined whether it is a normal area or a failure area, and is recorded in the normal / failure information recording unit 8 as normal / failure information. This normal / failure information is used in failure system determination described later.

FIG. 3 is a waveform diagram for explaining the operation of the signal feature quantity computing units 6-1 and 6-2 and the feature quantity comparison unit 7. First, the audio signal is divided into block (small area) units, for example, 1 block = 128 samples. Next, the feature amount in the block is calculated and extracted in each of the systems A and B. The feature quantities in the blocks of the systems A and B are represented by {f _Ai (p)} and {f _Bi (p)} (i = 1, 2,..., N), where the block position is represented by p. Each is represented. Here, N represents the number (type) of feature quantities to be extracted. As described above, various features such as an average value and a standard deviation of an audio signal can be used.

The difference between the feature quantities {f _Ai (p)} and {f _Bi (p)} between the two systems A and B is less than or equal to a predetermined threshold Th _i (that is, ^∀ i → | f _Ai (p) −f _Bi (p ) | ≦ Th _i ), it is determined that no fault has occurred in any system in the section, and the difference exceeds a predetermined threshold Th _i (ie, す な わ ち_i → | f _Ai (p) −f _{In the case of Bi} (p) |> Th _i ), it is determined that a failure has occurred in any system in that region. The determination result is recorded in the normal / failure information recording unit 8 for each block as normal / failure information.

  As described above, the signal feature quantity calculation units 6-1 and 6-2 and the feature quantity comparison unit 7 use global feature quantities that are not easily affected by coding noise and the like, and the difference values of the feature quantities between the systems. Is the norm of fault detection. Therefore, when the encoders are inserted in two systems, the encoding noise is generally different in the two systems, and this noise difference becomes a problem in fault detection by simple comparison between the two system signals. In the embodiment, the difference in coding noise is not erroneously detected as a failure, and sound quality degradation peculiar to a transmission error in which a broken portion exists can be accurately detected.

  Next, a configuration subsequent to the feature amount difference calculation units 9-1 and 9-2 will be described. This configuration is for fault system determination, that is, to detect which of the systems A and B has a fault. When the characteristic amount in each system shows a clear abnormal value, the system in which the failure has occurred can be determined by the threshold comparison units 14-1 and 14-2 and the silence / mute detection unit 15 described later.

  The feature amount difference calculation units 9-1 and 9-2 receive the feature amounts extracted by the signal feature amount calculation units 6-1 and 6-2, and the normal / failure information recorded in the normal / failure information recording unit 8 With reference to the information, the difference D of the feature amount between the matched region where the feature amount is matched between the systems A and B and the unmatched mismatched region is calculated for each system.

  The ‖D‖ comparison unit 10 compares the feature value difference D of the system A and the feature value difference D of the system B. The fault system determination unit 11 determines a faulty system based on the comparison result of the ‖D‖ comparison unit 10. A large feature amount difference D means that the audio signal has changed greatly between the matching region and the non-matching region, and therefore there is a high possibility that a failure has occurred. Furthermore, how large the difference is determines the degree of whether the determination is significant. The fault system determination unit 11 performs the determination after comprehensively considering the above.

FIG. 4 is a waveform diagram for explaining the operation of the fault system determination unit 11. The determination of the failure occurrence system is performed in units of a plurality of, for example, four audio signal blocks (hereinafter referred to as frames). Referring to the normal / failure information recorded in the normal / failure information record 8, the feature amount difference D = {D _i between the matched region where the feature amount is matched between the systems A and B and the unmatched mismatched region } (I = 1, 2,..., N) is calculated for each system, and the system with the larger value D is determined as the system in which the failure occurred. This determination result is output for each frame.

  As described above, the fault system determination unit 11 actively uses the general property that a failure area due to a transmission fault is localized (especially this tendency is strong in the case of a digital transmission fault). If the faulty part in the duplex audio signal is obvious, the difference between the feature value of the normal region and the feature value of the failure region in each system is used as a standard, and the one with the larger difference value Determined as a faulty system. In this case, since the system in which the failure has occurred is determined by paying attention only to the difference in the feature amount between the normal area and the failed area, stable determination can be performed regardless of the system error rate.

Specifically, first, the sum of absolute differences at the boundary part of the normal / failure area is obtained for each feature amount. B _j (j = 1, 2,..., N) is a boundary between blocks in the frame, and g ₁ (b _j ) and g ₂ (b _j ) are functions representing the positions of two blocks sandwiching the boundary b _j. If the whole boundary of the normal / failure area is C, the intra-system difference D _Ai of the i-th feature amount in the system A can be calculated by the following equation (2). Here, n is the number of boundaries between blocks, and the number of boundaries is 3 when there are 4 blocks per frame as shown in FIG. The same applies to the system B.

Based on the in-system difference D _i of the feature amount calculated as described above, for example, ‖D‖ is calculated by the following equation (3), and the system with the larger value of ‖D‖ is determined as the faulty system. judge.

Here, w _i (i = 1, 2,..., N) is a weighting coefficient for each feature quantity i, and can be determined as follows, for example. That is, a preliminary experiment using a no-failure signal is performed to obtain a distribution of difference values between adjacent blocks for each feature quantity i, and the reciprocal of the standard deviation is set as a weight w _i for the feature quantity i.

  The determination result of the failure system determination unit 11 can be used as a control signal, and the switch 3 can be switched to the standby system side when it indicates an active system failure. In that case, if the type and number of feature quantities are optimized, it is possible to increase the correct rate of failure system determination and switch to the standby system, but there may still be erroneous determination. Therefore, in this embodiment, the determination processing for each frame is further acquired by the majority processing unit 12 a plurality of times, and the majority determination is taken to obtain a final determination result, thereby improving the determination reliability.

  In addition, the significance determination unit 13 is provided, and the determination result used for the majority vote in the majority processing unit 12 is only the determination result made significant by the following formula (4), for example, thereby improving the effectiveness of the majority process. Can do.

  If the majority process is performed according to the significance determination by the above equation (4), the determination result when the difference between the feature amount difference of the system A and the feature amount difference of the system B is small contributes to the majority decision with low reliability (insignificant). You can avoid it.

  In the present embodiment, the threshold value comparison units 14-1 and 14-2 and the silence / mute detection unit 15 are provided, and the feature amounts and audio signals calculated by the signal feature amount calculation units 6-1 and 6-2 are received. It is checked whether or not a predetermined condition is met, and this check result is used in fault system determination and significance determination.

  When the feature amount calculated by the signal feature amount calculation units 6-1 and 6-2 clearly shows an abnormal value, it can be determined that a failure has occurred in the system indicating the abnormal value. An obvious abnormal value occurs, for example, when an audio signal of a large level that cannot be normal or a signal of 0 level continues. The threshold comparison units 14-1 and 14-2 provided for the respective systems A and B detect that the feature values of the respective systems A and B are abnormal values. The silence / mute detection unit 15 detects whether an abnormality in the audio signal level occurs only in one system. Here, when the signal level of only one system is always 0, there is a high possibility that the previous stage encoder, decoder, etc. of that system has caused a mute failure.

  The fault system determination unit 11 determines that a fault has occurred in one system when the feature value or signal level of one system indicates an abnormal value. Further, since the determination result of the fault system determination unit 11 when the feature amount or the signal level shows an abnormal value is considered to be highly reliable, the significance determination unit 13 increases the significance at that time.

  FIG. 5 is a waveform diagram illustrating an example of an audio signal including a failure area due to a failure. Here, an audio signal waveform including a failure area where the signal level has become extremely high due to a failure and a failure area due to a mute failure is shown. As described above, the failure area due to the transmission failure is localized.

  Note that the mute failure is a state in which a silence signal (0 level signal) is output as an audio signal as a result of detection of a transmission path failure by an encoder, a decoder or the like installed in the previous stage of failure detection processing. means. In such a case, if attention is paid to only one system, it cannot be distinguished whether the current audio signal is originally silent or a mute failure.

  However, in the present embodiment, the silence / mute detection unit 15 receives the audio signals of both systems, and does not determine that there is a failure when both systems are silent, but only when one of the systems is silent. Therefore, the reliability of fault detection is high.

  In the above embodiment, an encoder and a decoder are arranged in each system and the audio signal is digitally compressed and transmitted. However, compression encoding at the time of transmission is not always necessary, and the present invention The present invention can also be applied to a case where an audio signal is transmitted without being compressed.

  As described above, according to the present invention, it is possible to automatically detect a failure detection / failure system in real time and with high reliability by using a double-transmitted audio signal, and to perform duplexing. It is possible to contribute to non-interruption of transmission voice, that is, improvement of reliability.

It is a block diagram which shows one Embodiment of the failure detection apparatus of the double audio | voice signal which concerns on this invention. It is a figure which shows an example of the conversion characteristic of a signal logarithm conversion process part. It is a wave form diagram for demonstrating operation | movement of a signal feature-value calculating part and a feature-value comparison part. It is a wave form diagram for demonstrating operation | movement of a failure system determination part. It is a wave form diagram which shows the example of the audio | voice signal containing a failure area by a failure. It is a block diagram which shows the conventional duplex transmission apparatus.

Explanation of symbols

1-1, 1-2 ... encoder, 2-1, 2-2 ... decoder, 3 ... changeover switch, 4-1, 4-2 ... transmission path, 5-1. , 5-2... Signal logarithmic conversion processing unit, 6-1, 6-2... Signal feature amount calculation unit, 7... Feature amount comparison unit, 8. -1, 9-2 ... feature quantity difference calculation unit, 10 ... ‖D‖ comparison unit, 11 ... fault system determination unit, 12 ... majority decision processing unit, 13 ... significance determination unit , 14-1, 14-2... Threshold comparison unit, 15... Silence / mute detection unit

Claims (6)

In a fault detection device for a duplicated audio signal transmitted through two lines,
Feature quantity extraction means provided corresponding to each system and extracting a feature quantity for each small region of the audio signal of each system;
Feature quantity comparison means for comparing the feature quantities extracted by the feature quantity extraction means between systems;
A failure detection apparatus for a duplicated audio signal, comprising: a determination unit that determines that a failure has occurred in any of the systems based on a comparison result of the feature amount comparison unit. In a fault detection device for a duplicated audio signal transmitted through two lines,
Feature amount extraction means provided corresponding to each system, and extracting a feature amount for each small region of the audio signal of each system;
When a fault has occurred in any of the systems and the fault area in the duplex audio signal is clear, a calculation means for calculating the difference between the feature quantity of the fault area and the feature quantity of the normal area in each system,
A fault detection apparatus for a duplicated audio signal, comprising fault system determination means for determining that a fault has occurred in a system having a larger difference calculated by the calculation means. The failure system determination means uses a plurality of determination results based on the difference calculated by the calculation means, performs a majority decision thereof, and makes a final determination. At this time, the difference calculated by the calculation means is predetermined. 3. The fault detection apparatus for a duplicated audio signal according to claim 2, wherein a determination result when the value is equal to or less than the value is made insignificant. Furthermore, a silence / mute detection means is provided for detecting whether or not the audio signal in the small area is silent and comparing the audio signals in the two systems to detect whether or not the silence is due to a failure. 4. The fault detection apparatus for a duplicated audio signal according to claim 1, wherein 5. The duplicated voice according to claim 1, further comprising an abnormal value detecting means for detecting whether or not the feature quantity extracted by the feature quantity extracting means shows an abnormal value. Signal failure detection device. Furthermore, for the input audio signal, the sign is stored or inverted, the absolute value of the part with a relatively small absolute value is increased, and the absolute value of the part with a relatively large absolute value is decreased monotonically increasing or decreasing 6. The apparatus for detecting a failure in a duplicated audio signal according to claim 1, further comprising conversion means for performing conversion with the conversion characteristics of:

Images