WO1998018295A1

WO1998018295A1 - Triggering a measurement process for evaluating the quality of audio and/or speech signals

Info

Publication number: WO1998018295A1
Application number: PCT/EP1997/005814
Authority: WO
Inventors: Bodo Teichmann; Michael Keyhl; Heinz GERHÄUSER; Christian Schmidmer; Ernst Eberlein
Original assignee: Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V.
Priority date: 1996-10-21
Filing date: 1997-10-21
Publication date: 1998-04-30
Also published as: DE19730716A1; ATE184444T1; AU714063B2; JP3155283B2; KR20000048615A; DK0891685T3; EP0891685A1; CA2264112C; DE59700423D1; JP2000505269A; AU4948897A; EP0891685B1; KR100319465B1

Abstract

A process and device (24) for triggering a measurement process for evaluating the quality of audio and/or speech signals enable field measurements to be carried out to determine the quality of a digital transmission. During a first step, digital values of a reference signal underlying at least part of a test signal whose quality is to be evaluated are stored. Digital values of the test signals are also at least temporarily stored. The degree to which a section of the reference signal values coincides with the test signal values is determined to trigger the measurement process as soon as a predetermined degree of coincidence is exceeded, in order to compare a further section of the reference signal values to the test signal values in the relative position in time of the reference signal and test signal values at the moment when a predetermined degree of coincidence occurs.

Description

Triggering a measurement procedure for quality assessment of audio and / or speech signals

description

The present invention relates to the quality assessment of data-reduced audio and / or voice signals and in particular to a method and an apparatus for triggering a measurement method for quality assessment of audio and / or voice signals.

Modern methods for quality assessment of data-reduced audio and / or voice signals, such as NMR, PAQM or POM signals, do not use special, synthetically generated measurement signals for the measurement, but real audio and / or voice signals. This requires that the original signal and the signal to be assessed must be available for the measurement, ie for the quality assessment of audio and / or voice signals. The original signal is, for example, an audio and / or voice signal recorded in a recording studio, which is to be transmitted to a receiver by means of a cable connection or by means of a wireless radio link. In order for an economical transmission to be possible, this signal must be data-reduced or encoded, after which it is fed as a bit stream to a digital transmitter which, for example, carries out pulse amplitude guadrature modulation, as a result of which the data-reduced audio and / or voice signal, ie the bit stream, is modulated onto a carrier . This carrier is emitted by means of a suitable antenna in the case of a wireless radio transmission and received and demodulated by means of a suitable receiving antenna. After demodulation of the bit stream modulated by a suitable modulation method, it is similar to the bit stream present in the receiver before the modulation, apart from influences of the radio link. After inverse coding or decoding and after a digital / analog conversion, the signal originally recorded in the recording studio can again be made audible.

Methods for quality assessment of data-reduced audio signals determine differences in the audio signal made audible in the receiver with respect to the audio signal recorded in the recording studio in the present example, in order to detect either coding or decoding errors or bit errors introduced by the transmission path. A reference signal, i.e. the signal recorded in the recording studio, with a test signal, i.e. compared to the signal made audible in the receiver after the digital / analog change, in order to assess the quality of the coding, the radio link and the decoding.

As already noted, it is necessary that the reference signal and the test signal must be available during the measurement. The delay between these two signals must be less than 500 ms for current measurement methods. This is not a problem under laboratory conditions, which is why the quality assessment of data-reduced audio and / or speech signals is already carried out under laboratory conditions.

However, if field measurements need to be performed, there is normally no reference signal available because the signal source, i.e. the recording studio or a transmitting antenna is arranged spatially far from a measuring point.

It is obvious that in order to assess the quality of audio signals, the reference signal on which the test signal is based must be known. Due to the fact that it is not necessary for the quality assessment to constantly compare the test signal with the reference signal, it is possible, assuming changes in the coding or decoding or the radio link that are not too great, only a section of the test signal, which is a reference signal to compare with the reference signal. Of course, this section of the test signal, the the reference signal is based, and the reference signal itself may be present in the measuring device. Constantly recurring identical audio signals from radio programs are, for example, jingles which initiate certain radio programs and are not changed over long periods of time, since they make a significant contribution to identifying a listener with "his" radio program. It is thus possible to carry out quality measurements of a received audio and / or voice signal by providing a jingle at the transmitter at many different locations and at many different times.

Methods for jingle recognition are also already known. However, these methods only serve to recognize a specific music title, e.g. for survey purposes or for monitoring correct billing of radio stations, the purpose of which is merely to recognize whether a certain music title has been played or not. These known methods for jingle recognition only provide a measure of conformity with a specific signal at the end of the jingle. However, this output is unusable for a measuring method for quality assessment of audio and / or voice signals, since the measuring period has already ended when the jingle has been recognized. Likewise, in the known methods for jingle detection, an exact compensation of the delay between the test signal and the reference signal is not possible.

The object of the present invention is to provide a precise triggering of a measuring method for the quality assessment of audio and / or speech signals.

This object is achieved by a method for triggering a measuring method for quality assessment of audio and / or voice signals according to claim 1 and by a device for triggering a measuring method for quality assessment of audio and / or voice signals according to claim 9. The invention is based on the knowledge that a jingle is particularly suitable before the start of a radio broadcast in order to be able to be used for quality assessment of the radio signal. The prerequisite for this is that such a jingle is sent during the measurement period. The reference signal, ie the jingle, before coding and sending, can be recorded at the transmitter before the measurement and then brought to the receiver in any way, such as by modem or magnetic tape. A jingle detector according to the present invention continuously compares the received signal with the provided or stored reference signal, only a small section of the reference signal being compared with the test signal during the continuous comparison, and after reaching a predetermined degree of agreement at least a further section of the Reference signal is compared with the corresponding section of the test signal. As soon as a match has been detected, the jingle, ie the reference signal, is played in synchronism with the test signal on which the jingle is based since the start of the jingle has been identified, after which any measuring method known in the art for comparing the two signals is used to Quality assessment of coding, decoding, transmission, etc. can be used.

A preferred embodiment of the present invention is explained in more detail below with reference to the accompanying drawings. Show it:

1 shows a system arrangement which can carry out the present invention; and

Fig. 2 is a schematic block diagram of the device for triggering a measuring method for quality assessment of audio and / or voice signals.

1 shows a schematic system arrangement which outlines the method and the device of the present invention. can lead. An audio source 10, which can be a recording studio or a magnetic tape recorded by a recording studio, is connected to a transmitter 12, which in turn is assigned a transmission antenna 14. The transmitter 12 can carry out an analog / digital conversion of the audio and / or speech signal supplied by the audio source 10. The digital audio and / or voice signal is also encoded in the transmitter 12, ie data reduced, as is known to those skilled in the art and modulated onto a carrier by means of a suitable modulation method, which is also known to those skilled in the art, so that the encoded and modulated audio and / or voice signal can be emitted via the transmission antenna 14. The coded and modulated audio and / or voice signal propagates over a radio link 16 in order to be received by a receiving antenna 18. It is obvious to a person skilled in the art that the radio link 16 is only an example, since it can also be implemented by a cable connection, which can transmit digital radio signals or digital television signals, ie multimedia signals, for example.

The coded and modulated audio and / or voice signal received by the receiving antenna 18 is demodulated and decoded in a suitable manner in a receiver 20 and converted to analog / digital and is fed via a receiving line 22 to a jingle detector 24 which is used to trigger a measuring method Represents quality assessment of audio and / or voice signals and is connected via a detector connection 26 to a measuring device 28 (Fig. 2), which can perform one of a variety of measurement methods known to those skilled in the art for quality assessment of audio and / or voice signals.

A dashed line from the connection of the audio source 10 and the transmitter 12 to the jingle detector 24 represents a reference signal line 30. The dashed line of the reference signal line 30 is intended to symbolize that this connection between the audio source 10 and the jingle detector. gate 24 does not have to be a continuous connection since, as has already been noted, a reference signal, ie the audio signal of a jingle, has to be supplied to the jingle detector 24 only once or at certain times before the coding and modulation. In contrast to this, the solid connections represent constantly active data connections, at least during a measurement process for the quality assessment of audio signals.

FIG. 2 shows a more detailed illustration of the jingle detector 24, which in a preferred exemplary embodiment can be a personal computer with suitable software and with a suitable operating system. The jingle detector 24 has a memory device 242 for storing the reference signal and for storing the test signal, which is fed to a digital signal processor card 244 via a test signal input 22. If necessary, the test signal can be fed to the measuring device via a first output 26a, while a second output 26b of the digital signal processor card 244 will feed the reference signal to the measuring device 28 if required. A communication line 26c is used for initialization tasks and for measuring communications between the jingle detector 24 and the measuring device 28. The fact that the communication line 26c is shown in dashed lines, in analogy to the reference signal line 30 from FIG. 1, means that the communication line 26c is not continuously active , but only signal the start or end of a measurement by the measuring device 28, for example.

In a preferred embodiment of the method of the present invention, the reference signal, suitably transmitted from audio source 10 to jingle detector 24, was recorded on storage device 242, which is a hard disk of a standard personal computer. The standard personal computer has, as additional hardware, the digital signal processor card 244 in the form of a plug-in card which is used for the audio test Signal input 22 and the first output 26a and the second output 26b. At this point it should be pointed out that the three lines 26a to 26c, as can already be seen from the reference numerals assigned to them, are contained in the detector connection 26 from FIG. 1. The test signal input 22 of the signal processor card 244, which can contain several signal processors, is, as can be seen from FIG. 1, suitably connected to the receiver 20. The first output 26a and the second output 26b of the digital signal processor card 244 are connected to corresponding inputs of the measuring device 28 for the test signal or for the reference signal.

According to a preferred exemplary embodiment of the method of the present invention, a first section of the values of the reference signal is read from the hard disk 242 into a memory of the digital signal processor card 244 at the beginning of a measuring cycle. This first section can be about 1.5 seconds long. This first section of the values of the reference signal is filtered in the digital signal processor card 244, subsampled and then compared by means of a cross-correlation with a section of the test signal which is treated in the same way and which is supplied via the reception line 22 to the digital signal processor card 244 or the memory device 242. The test signal was also filtered and subsampled.

It will be apparent to those skilled in the art that analog / digital conversion of the test signal can be performed in the jingle detector 24 by a suitable section of the digital signal processor card 244 if the receiver 20 outputs analog values. However, if the receiver 20 outputs digital values of the test signal, then an analog / digital conversion in the jingle detector 24 is not necessary. It is also apparent that the cross-correlation between a portion of the values of the reference signal and a portion of the values of the test signal is performed digitally. In a preferred embodiment the algorithm for performing the cross-correlation, ie the device for determining the degree of correspondence of a section of the values of the reference signal with the values of the test signal, is carried out in software. In a departure from this exemplary embodiment, the device for determining the degree of correspondence of a section of the values of the reference signal with the values of the test signal can also be implemented in terms of hardware by means of suitable shift registers and logic gates, which essentially carry out an exclusive-OR operation .

As already mentioned, the comparison between a section of the values of the reference signal and a section of the values of the test signal to be assessed takes place section by section. In a preferred embodiment of the present invention, a section consists of 512 samples of the test signal, this section also being referred to as a "frame". If the portion of the values of the reference signal is made shorter, the comparison operation will run faster, but the number of incorrect detections of the correspondence between the test signal and the reference signal will increase. If the portion of the values of the reference signal that is used to determine a degree of correspondence is chosen to be longer, the time required for this operation increases, but the probability of an incorrect detection of a correspondence will decrease.

If the degree of correspondence or degree of correlation exceeds a certain value, the method for triggering a measurement method for quality assessment of audio signals determines that the values of the test signal which have just been compared with the section of the values of the reference signal represent the beginning of a jingle, which as a reference signal in its original form is present. If the predetermined degree of agreement or the predetermined degree of correlation is exceeded, a measuring method is used Triggered quality assessment of audio signals. The jingle detector 24 notifies the measuring device 28 of the start of the measurement via the communication line 26c, whereby the measuring device 28 is activated in order to receive a further section of the values of the reference signal via the first output 26a of the jingle detector 24, essentially simultaneously via the second Output 26b of Jingledetector 24 the corresponding values of the test signal are transmitted to the measuring device. After triggering the measuring method, which is carried out by the measuring device 28, further sections of the reference signal are read from the hard disk 242 and output together with the test signal received via the receiving line 22, ie transmitted to the measuring device 28. The first output 26a and the second output 26b can be four-channel audio interfaces or, in the case of stereo outputs, the left channel and the right channel of the digital audio signal. The left channel can then transmit the test signal, which corresponds to output 26b, whereas the right channel, ie output 26a, will transmit the reference signal from hard disk 242.

By correlating the section of the values of the test signal with a corresponding number of values of the reference signal, the delay between the reference and the test signal is automatically eliminated, since the section of the values of the reference signal is compared with the continuous bit stream of the received test signal. A new section of values of the reference signal, which is transmitted from the jingle detector to the measuring device, corresponds to the values of the test signal on which the reference signal is based, since the transmission of further sections of the test signal to the measuring device only starts when the jingle detector assumes that that the currently existing section of the values of the test signal which was just used for the correlation evaluation is the section which is based on the section of the values of the reference signal used for the correlation determination. In contrast to the above-described jingle detector according to the prior art, in one exemplary embodiment of the present invention the entire jingle minus the first section of the values of the reference signal is available for quality assessment for the measuring device. In the prior art, the entire jingle must first have been played in order to then be able to determine that the current test signal was a jingle.

Since the delay between the two is known due to the correlation of the two signals, the delay being, in principle, less than a section or frame length, this delay can also be compensated for. The first section of the reference signal used by the jingle detector 24 to determine the correspondence can also be provided for the measurement method for quality assessment of audio signals if both signals are output simultaneously from the digital signal processor card.

It should be noted that in the exemplary embodiment described so far, the reference signal is stored continuously on the hard disk 242, whereas the test signal is only temporarily stored for correlation. After determining the correspondence between the section of the values of the reference signal and values of the test signal, further received values of the test signal are not temporarily stored in the jingle detector 24, but are output directly to the measuring device 28 via the second output 26b. However, if the first section of the values of the reference signal is also to be used for the measurement, this is only possible if both signals are output simultaneously from the digital signal processor card 244. If the digital signal processor card 244 has sufficient working memory, a longer section of the test signal can also be kept in memory and delayed. When the jingle is detected, the entire jingle and not just the part of the jingle that is sent after the detection can be measured become .

As has already been noted, this phase represents the actual measurement phase that is carried out in the measuring device 28. The only function of the jingle detector 24 in this measuring phase is to transmit further sections of the values of the reference signal to the measuring device 28 in synchronism with the received values of the test signal. It is obvious to a person skilled in the art that the number of values of a section in the measuring phase must be the same for the test signal and the reference signal. It will also be apparent to those skilled in the art that the reference signal can consist of any number of sections, depending on what a user desires. However, the reference signal must be divided into at least two sections, in which case a section from the jingle detector 24 is used to determine the correspondence between a corresponding section of the test signal and the first section of the reference signal. The further section is then used by the measuring device 28 with a corresponding section of the test signal for quality assessment of the test signal. As described above, however, the first section of the reference signal and the corresponding section of the test signal can also be used for quality assessment by the measuring device 28 by temporarily storing a longer section of the test signal.

If by suitable software in the jingle detector 24 before the start of the measurement phase, i.e. the phase of the transmission of the test and reference signals via the outputs 26b and 26a, and after this phase no signal or a very small signal is output, the measuring device 28 can be programmed in a manner known to those skilled in the art so that the beginning and the end of the measuring phase can be recognized automatically. The beginning and the end of the measuring phase are transmitted from the jingle detector 24 to the measuring device 28 via the communication line 26c.

Claims

claims

1. Method for triggering a measuring method for quality assessment of audio and / or voice signals, with the following steps:

providing digital values of a reference signal which is based on at least part of a test signal to be assessed in terms of its quality;

at least temporarily storing digital values of the test signal;

Determining the degree of agreement of a portion of the values of the reference signal with the values of the test signal;

Triggering the measurement process as soon as a predetermined degree of agreement is exceeded.

2. The method according to claim 1,

in which the triggering of the measuring method, as soon as a predetermined degree of agreement is exceeded, for comparing a further section of the values of the reference signal with the values of the test signal in the temporal relative position of the values of the reference signal and of the test signal when the predetermined degree of agreement occurs .

3. The method according to claim 1 or 2,

in which the part of the test signal to be assessed in terms of its quality, on which the reference signal is based, is a jingle which is repeated regularly or irregularly in the test signal.

4. The method according to any one of claims 1 to 3, in which the test signal is a decoded digital broadcast signal, while the reference signal is a section of the digital broadcast signal before it has been encoded, transmitted and decoded.

5. The method according to any one of the preceding claims,

in which the reference signal and the test signal are undersampled and filtered before being stored in order to obtain digital values of the test signal and the reference signal, respectively.

6. The method according to any one of the preceding claims,

in which the degree of correspondence between a section of the values of the reference signal and the values of the test signal is determined by means of a correlation.

7. The method according to any one of the preceding claims,

at which after the predetermined degree occurs

Match more than another section of the

Values of the reference signal are compared with corresponding values of the test signal.

8. The method according to claim 7,

in which, before the occurrence of the degree of correspondence between the values of the first section and the values of the reference signal and the subsequent first occurrence of a smaller than the predetermined degree of correspondence, a respective signal is generated which indicates a beginning or an end of the measuring method for quality assessment of audio and / or displays voice signals.

9. Device (24) for triggering a measuring device (28) which carries out a measuring method for quality assessment of audio and / or speech signals, with the following features:

a first memory device (242) for storing digital values of a reference signal which is based on at least part of a test signal to be assessed in terms of its quality;

a second memory device (244) for at least temporarily storing digital values of the test signal;

means (24) for determining the degree of correspondence of a portion of the values of the reference signal with the values of the test signal;

a trigger device (24) for triggering the measuring method carried out in the measuring device (28) as soon as a predetermined degree of correspondence is exceeded.

10. The device (24) according to claim 9,

in which the first storage device (242) is a hard disk, while the second storage device (244) is a working memory of a digital signal processor card.

11. The device according to claim 9 or 10,

in which the device for determining the degree of correspondence and the trigger device are present in a digital signal processor card (244), which has an input (22) for the test signal and a first - solve

most output (26a) for the reference signal and a second output (26b) for the test signal, the first and second outputs (26a, 26b) being connectable to a measuring device (28) for assessing the quality of audio signals.

12. The device according to any one of claims 9 to

in which the jingle detector (24) has a communication line (26c) which can be connected to the measuring device (28) in order to inform it of a start or an end of a measurement.