EP0598682A1 - Method for evaluating behaviour of radio listeners, and apparatus therefor - Google Patents

Abstract

To evaluate the audience behaviour to radio or TV programmes, listening samples are taken at particular times. These listening samples can detect the sound impressions actually received by a test person, on the one hand. For this purpose, the test person carries a portable monitor on his person. Furthermore, the receiving sets themselves can also carry such monitors, in which arrangement, apart from the acoustic pick-up of the listening samples, a direct electrical connection to the monitor is also possible. The listening samples are converted into digital form, compressed, transformed and, if necessary, coded by adequate methods. The listening samples are transmitted at relatively large time intervals to a central station where the listening samples are correlated with samples from the monitored programmes which are picked up at the same time. <IMAGE>

Description

The present invention relates to the field of radio listener behavior research, and more particularly to a method for determining the particular program being listened to.

A wide variety of methods are known in the field of television audience research in order to ascertain the program received and reproduced in each case and the identity of the viewers who are watching the program being reproduced. The possibility of identifying program parts recorded on video recorders and reproduced at a later date was also realized.

Although obvious and in principle also usable, these methods are impractical in the field of radio. One reason for this is that each television set is monitored, which makes sense from the point of view that there is a small number of television sets per household (usually one or two) that have a fixed location. Then relatively complex monitoring devices that can be adapted to the respective device can be attached.

In contrast, there are usually several radio sets in one household, some of which are also used outside the home. These include, for example, car radio, Walkman, portable radio. In addition, radio is often listened to in department stores, supermarkets, restaurants, etc. In addition, the frequency ranges of the radio programs are not nationwide. In order to determine the received program from the setting of a radio device, it would therefore be necessary to also determine the current location, and when evaluating a large and constantly updated one List of regional program broadcast frequencies to determine the program being played.

Monitoring the playback of the devices and the determination of the received radio program from the setting of the receiver, which is common for television receivers, is therefore not meaningful for radio receivers.

It is an object of the present invention to provide a method for determining the radio listener behavior, which allows the radio program heard in each case to be determined independently of the location of the listener.

A method that solves this problem is specified in claim 1. The further claims contain preferred embodiments, devices for carrying out the method and applications of the method.

Accordingly, acoustic hearing samples are recorded and stored in the vicinity of a listener in order to correlate them with corresponding recordings of one or more programs at a later time. For correlation, information about the point in time is also stored with the audio samples, or the audio samples are recorded in a predefined time grid, the start time of which is defined or z. B. is stored at the beginning of the audio sample recording. In order to reduce the memory required for the audio samples, it is advantageous to record and save at specific times and only over a short time interval.

The corresponding device, hereinafter referred to as the monitor, can be attached to the radio receiver and can also be carried by the test person. The former variant presupposes that it is also determined whether and who is listening to the radio, that is to say that it is sufficiently close. This can be done by input on the monitor, such as. B. by means of a keyboard, a code card or a barcode reader, or also by code transmitter that are carried by the people in the test household.

The second variant, the monitor carried by the test person, makes it easy to determine what the wearer of the monitor is hearing. A prerequisite for this procedure is a small and light monitor, which, for. B. can also be worn in a shirt breast pocket. The monitor must be functional for a sufficiently long time. On the one hand, this requires a sufficiently high storage capacity, and on the other hand, because of the mobile operation, a power supply via batteries, possibly supported by solar cells or other energy sources accessible in mobile operation. In order to use these energy sources as economically as possible, the audio samples are recorded exclusively on storage media that have no mechanical components, in particular no magnetic tapes, etc.

However, the applicable storage media, such as non-volatile or battery-buffered semiconductor memories, have only limited capacity and can generally only store digital data. Suitable procedures for the most compact possible storage and reduction of the data volume are therefore necessary. The recording of only short samples, interrupted by relatively long time intervals, already considerably reduces the amount of data to be saved. A further reduction can take place by restricting it to a reduced frequency band within the range of the audible frequencies, the audio band.

After the analog-to-digital conversion, further compression methods can be used, which can be carried out by a processor. These include those from digital technology known, generally applicable data compression methods, as well as transformations, such as the (fast) Fourier transformation. The degree of data reduction can be set here by the density of the support points.

Another aspect is data and privacy protection. It is foreseeable that test persons will not be happy to carry a device with them, or even have it in their environment, which among other things can listen to and record every spoken word. In the method according to the invention, protection against this is already provided by storing only short hearing samples. The other processing steps mentioned, such as reduced monitored frequency band and the use of transformations, further complicate the tracking of a call. Finally, the data can be additionally encrypted before being stored, preferably using a non-reversible method.

The program heard is determined by comparing the audio samples with program samples recorded at the same time. Program samples can e.g. B. in the broadcasting studio, at any point in the broadcasting system or in a cable reception system, where the assignment of channels to a program is defined. This does not have to be done via an acoustic intermediate stage, but can be done in any way. It is essential that the audio samples are recorded overlapping in time with the program samples, and that the recorded data can be brought into a form in which they can be checked for correspondence by a correlation method. The program samples can immediately be subjected to the same procedures as the audio samples. Another possibility is that the program samples are first saved in a raw version and only at the moment of correlation to the audio samples be adjusted. Program samples can thus also be used for correlation with stored audio samples that have been compressed, transformed and / or encrypted in various ways.

The prerequisite is that the time at which the audio samples and the program samples were recorded can be determined at the time of the correlation. The easiest way to do this is to save the beginning of the recording together with the samples. In the case of a series of samples recorded one after the other, it may also suffice to save the absolute time only once and later only to record the time intervals between two samples, or to calculate the times of the further samples from the saved time assuming a regular sample recording .

The correlation of the audio samples with the program samples takes place in a central office. The transmission of the audio samples from the test households to the headquarters can be done in different ways. Because of the small size of the monitors, they can be sent directly to the headquarters by post. A device can also be installed in the test households that can read out the memory of the monitors and transmit the data to the control center via modem. A system for determining television habits may already be installed in such a test household. Such systems often already include a modem with a data connection to a central office. About this system can then. B. the data of the monitors can also be transmitted to the control center using a special adapter.

The program samples can be recorded by devices similar to the monitors. Since these devices are stationary in most cases, the program signal is directly decoupled, i.e. without detour via acoustic reproduction, applied, and it can also be recorded during longer recording intervals, since the program samplers can contain more expensive storage media of higher capacity. So z. B. time shifts between audio samples and program samples, as they can result from time bases that are not running exactly synchronously, are subsequently compensated for during the correlation without the effective correlation interval being shortened.

Figur 1

zeigt anhand eines Blockschaltbilds das Funktionsprinzip, und

Figur 2

zeigt ein mehr in Detail gehendes Blockschaltbild.

The invention is to be further explained in an exemplary embodiment with reference to figures.

Figure 1

shows the principle of operation using a block diagram, and

Figure 2

shows a more detailed block diagram.

From the microphone 1, the audio signal reaches the bandpass filter 2, which filters a frequency range from 100 to 4000 s⁻¹, preferably 300 to 3000 s⁻¹, from the entire audio frequency range. The bandwidth-limited signal is passed on to the analog-to-digital converter 3 and from there further in digital form to a first buffer memory 4. The analog-to-digital converter is controlled by a timer 9. This timer turns on the analog-to-digital converter for one second at the beginning of every full minute. For this purpose, the timer 9 has a quartz time base which is synchronized with a time base in the control center.

The audio sample stored in digital form in the memory 4 during the one-second recording time is read out from the memory 4 in the recording-free rest of the minute and after Fourier transformation in the Fourier transform unit 5, compression in the compressor 6 and encryption in the encryption unit 7 in the memory 8 filed.

Conceivable embodiments are, for example, as a check card or clip or also integrated in a watch.

The monitors in the test household are placed once a day in a charging and synchronization station that is once available in the household. The synchronization station ensures that the timers 9 synchronize in the individual monitors. Furthermore, it can read the memory 8 from each monitor and stores the content in its own, larger-sized memory, which, for. B. can be a tape deck. The memory of the monitor is then expediently deleted. As a further function, the synchronization station can charge rechargeable energy sources contained in the monitor.

The data transfer to the control center can be done by e.g. B. the charging and synchronization station is called by the central computer at night or it calls the central computer, after which it transmits the stored data on request. The timer present in the charging and synchronization station can also be brought into agreement with the time base of the central computer. The connection to the control center is established either via its own modem or via the modem of an existing system for researching audience behavior. The second option is to send the data carrier of the loading and synchronization station to the control center. Finally, the monitors themselves can also be sent to the control center, whereby the built-in timer can in turn be synchronized with the time base of the control center.

The monitor can be adapted to the current conditions in a variety of ways. In particular, the various processing procedures can be carried out by only one processor. With a recording time of just one Seconds per minute he has 59 seconds to perform the transformation, compression and encryption. Performing all of these operations by a processor has the advantage that a change in the overall functioning is only possible by exchanging the program that controls the processor. For example, the encryptor or its key can be exchanged, the compression can be switched on and off, and other transformation methods can be used.

The filter 2 can also be designed digitally and then arranged behind the analog-digital converter. The filtering can also be carried out in a digital manner by a processor. The order of the functional elements can also be chosen differently overall.

The synchronization of the time base 9 can be dispensed with in that the time base 9 has a receiver for a time normal signal, e.g. B. DCF-77, which is also used by the program samplers as a time standard.

A problem with the purely acoustic recording of the audio samples is that the ambient noise level may be too high. A possible solution is that so-called mini spies are attached to the speakers, which pick up the structure-borne noise of the respective speaker and radiate it directly to the monitors, e.g. B. by means of radio signals, infrared or ultrasound. The monitor can also be connected directly to the speaker or headphone output. B. a Walkman can be connected. A monitor can also be attached to a hi-fi stereo system using an adapter that allows the listener to be input. The monitor could also be equipped with a button that allows the wearer to deactivate the monitor, e.g. B. during a confidential conversation.

A preferred embodiment will now be described with reference to FIG. 2. Here, the monitor, that is, the hearing sampler that the listener carries with him, is integrated in a watch, preferably a wristwatch. The microphone 1 has a high to very high sensitivity for the sound waves 11 within a frequency range of approx. 200 s⁻¹ to 3000 s⁻¹, but outside of this range a low one. As a result, the microphone 1 performs pre-filtering. The branch 13, which is used for feeding signals instead of or in addition to being recorded via the microphone 1, is discussed below.

From the microphone 1, the already bandwidth-limited (approx. 200-3000 s⁻¹) signal 14 arrives at an amplifier and bandpass 15, which raises the signal for the subsequent A / D converter 3 to the necessary level and the bandwidth limitation with the desired sharpness causes. The output signal 18 of the amplifier 15 reaches the A / D converter 3, which converts it into a digitized form 20. The conversion takes place with a sampling rate of 6000 s⁻¹ and a resolution of 8 bits (= 1 byte) per sample. Switching through the output signal during the sampling interval is carried out by switch 22. Since the sampling interval lasts one second, 6,000 values (= 6,000 bytes) reach the buffer memory 4 in the meantime. Switches 22 and A / D converter 3 are controlled by the timer 9, which can be coupled to the time display of the clock or works separately from it. the timer 9 the A / D converter 3 provides the sampling rate signal 24 of 6000 s⁻¹ and the switch 22 the switching signal 25 with a duration of 1 s per minute.

The data 29 are read out from the buffer memory 4 by the Fourier transform unit 5. The Fourier transform unit processes 200 ms sections of the one-second samples, so that the result per sampling interval five Fourier coefficient sets 30 arise, which are initially stored in a buffer 32.

From this, the Fourier coefficients 34 are fetched from the data compressor 6 and reduced in scope without loss of information by one of the common compression methods, such as Huffmann coding, Lempel-Ziv-Welch method, etc. Compressions of up to 2/5 of the initial data volume can be achieved here, or even more if more powerful methods are known and used. An empirical value is that approx. 100 bytes of compressed data per sampling interval, i. H. per minute. The compressed Fourier coefficient data sets 37 are stored in a non-volatile memory 8, e.g. B. a flash EPROM, with a capacity of z. B. 1 MB (= 2²⁰ bytes) saved. With this capacity, with a data flow of 100 bytes per minute, an uninterrupted watch wearing time of a few days until the memory is read out can be achieved. As the storage capacity continues to increase as technology continues, this time will increase.

At least the Fourier transformer unit 5 and the data compressor 6 are implemented by a processor under the control of a respective program. At the same time, if necessary, the processor can also take over the encryption mentioned above.

The reading out of the data from and the deletion of the EPROM 8 takes place in the loading and synchronization station 39, into which the watch containing the monitor can be inserted. The energy and data transmission takes place contactlessly via electromagnetic coupling, which avoids opening the watch and requires only a small amount of instruction from the operator to load and read out, so it can easily be carried out by the test person himself. The monitor in the watch indicates contact with the charging station 39 the functional elements readout control 40 and reset and erase control 41 for reading out and erasing the memory 8, a timer synchronization unit 42 for synchronizing the monitor timer 9 and a battery charger 43 which contains the supply battery of the monitor and is used to recharge this battery. The data transmission is preferably serial.

The synchronization of the clock 9 is necessary because the 200 ms divisions of the 1 second samples must be correlated exactly in time with the samples in the transmitter, which can also cause problems for days with a quartz time base. To compensate for one of the most important influencing variables, the temperature, the temperature dependence of the timer 9 is determined and stored for each monitor. With the help of a timer 9, d. H. in particular with the quartz, coupled temperature sensor contained therein, the time deviation can be determined and corrected by the influence of temperature, whereby an accuracy of ± 200 ms gait errors per week can be achieved.

Other possible additional devices are a person identification 45, a “applied” detector 46, an attention button 48 or the like for inputting that listening is being carried out with increased attention, etc. The person identification 45 is predefined as the identifier of the wearer. However, it is also conceivable to provide several identifiers from which the clock can be selected using operating elements.

The "applied" detector 46 serves to automatically determine whether the watch, ie the monitor, has been put on by the test person and thus the hearing samples also correspond to the auditory impression of the person. As a secondary function, the hearing monitoring can also be switched off when the watch is not on, e.g. B. at night. The detector 46 can be inductive or capacitive Be designed sensor that responds for example to the proximity to the body of the test person or the condition of the bracelet (open / closed).

The attention button 47 serves to allow phases of increased attention to be marked manually by the test person.

The comparative hearing samples, the transmitter side, z. B. recorded in the studio, are processed and saved by a functionally identical monitor. Instead of via the microphone 1, however, the signal is fed in via the branch 13 directly at the input of the amplifier 15. The final storage takes place on a data medium of large capacity as usual in computer technology, e.g. B. on the basis of magnetic tapes. The functions required for portable, automatic operation, such as the interface units for the charging and synchronization station, attention button, "applied" sensor etc. are also omitted.

The invention can also be used to determine participant behavior for other types of radiation having an acoustic component, such as, for. B. television.

Claims (16)

Method for identifying an auditory impression that is perceived by a listener, preferably a radio or television program, characterized in that audio samples are recorded and stored in the vicinity of the listener and that the audio sample is recorded at specific times and / or information about the audio sample the time of recording is stored, the recorded signal being digitized and stored in a non-volatile memory device (8) without mechanically movable components which are functionally important for the storage. A method according to claim 1, characterized in that radio programs perceived by a listener are determined. Method according to one of claims 1 or 2, characterized in that audio samples are determined by recording acoustic signals (11) at certain recording times without or with environmental noise during a recording interval, the recording interval being substantially smaller than the period between two recording times, whereby the recording time is 0.1 s to 10 s, preferably 0.5 s to 2 s. Method according to claim 3, characterized in that the recording interval is divided into sub-intervals and the data of the hearing samples for these sub-intervals are determined and stored independently of one another. Method according to one of claims 1 to 4, characterized in that a frequency band from about 100 s⁻¹ to 4000 s⁻¹, preferably from about 300 s⁻¹ to from the hearing sample about 3000 s⁻¹, or a portion of it is filtered out. Method according to one of claims 1 to 5, characterized in that the recorded signal is subjected to a transformation, in particular a Fourier, Fast Fourier or Laplace transformation, and / or compressed. Method according to one of claims 1 to 6, characterized in that the storage takes place on non-volatile or battery-buffered semiconductor memories, in particular electrically erasable read-only memories. Method according to one of claims 1 to 7, characterized in that encryption, in particular non-reversible encryption, is applied to the digitized signal before storage. Device for carrying out the method according to one of claims 1 to 8, characterized in that it has at least the following components, which are connected to one another in the specified sequence via output or input: a microphone (1); at least one component (1, 15) filtering out a frequency band; an analog-digital converter (2); a processing unit (5, 6, 7; 5, 32, 6) for the digitized data, in particular including a processor which can carry out the processing under the control of a program; and a non-volatile memory (8) for digitized data without mechanically movable components essential for the storage process, in particular a non-volatile or battery-buffered semiconductor memory. Apparatus according to claim 9, characterized in that the non-volatile memory is an electrically erasable Semiconductor read-only memory is, preferably of the FLASH-EPROM type. Apparatus according to claim 9 or 10, characterized in that the timer (9) contains an absolute time base which can be set to a reference time base by means of a synchronization unit (39), and the absolute time base is provided with at least one correction unit for time deviations caused by temperature fluctuations, wherein the correction unit reacts to the temperature in the vicinity of the time base. Device according to one of claims 9 to 11, characterized in that the timer (9) contains a receiver and decoder for standard time signals. Device according to one of claims 9 to 12, characterized in that the device can be carried by the test person and has a capacitive or inductive sensor (46) for generating a signal which indicates that the person is being carried. Device according to one of claims 9 to 12, characterized in that a microphone, in particular a structure-borne sound microphone, is present on a loudspeaker of a radio or television receiver, a transmitter is connected to the microphone which radiates the signal picked up by the microphone, and that the device contains a receiver for this signal. Method for determining the listener or viewer behavior, characterized in that hearing samples are determined according to one of claims 1 to 8, that program samples, which at least partially overlap with the audio samples, are created and stored by at least one listening or television program, the Audio samples and the program samples go through the same processing steps between recording and storage, and that temporally corresponding audio samples and program samples are correlated at a later point in time, with a positive result of the correlation, there is agreement between the program recorded by the audio samples and the program samples. A method according to claim 15, characterized in that the program samples run through a first part of the processing steps prior to storage, and between processing read out from the storage medium and the correlation run through further processing steps, in particular encryption, in order to condition the program samples for correlation with the audio samples.

Images