System for broadcast of data in an audio signal by substitution of imperceptible audio band with data6151578Abstract A system for broadcasting data (D) that can transmit information in the passband of a broadcast audio-frequency signal (S). The system can determine at least one frequency band (F'.sub.13, . . . , F'.sub.24) and the amplitude (A'.sub.13, . . . , A'.sub.24) of the audio-frequency signal (S). The system compares this amplitude with an auditory masking level (Nm(13), . . . , Nm(24)) associated with this frequency band and eliminates the frequency components of the audio-frequency signal in the frequency band if the amplitude of the signal is lower than the auditory masking level of the band. The system can insert the data in this frequency band at a level lower than or equal to the auditory masking level of the frequency band. Claims What is claimed is: Description BACKGROUND OF THE INVENTION
______________________________________
Central frequency
Critical band fc (Hz) Bandwidth (Hz)
______________________________________
B1 60 80
B2 150 100
B3 250 100
B4 350 100
B5 455 110
B6 570 120
B7 700 140
B8 845 150
B9 1000 160
B10 1175 190
B11 1375 210
B12 1600 240
B13 1860 280
B14 2160 320
B15 2510 380
B16 2925 450
B17 3425 550
B18 4050 700
B19 4850 900
B20 5850 1100
B21 7050 1300
B22 8600 1800
B23 10750 2500
B24 13750 3500
______________________________________
We can see that the critical bands have variable widths, the narrowest being the first critical band B.sub.1, which covers the lowest frequencies, and the widest being the twenty fourth critical band B.sub.24 which covers the highest frequencies. For each critical band, standard ISO/IEC 11172-3 defines a critical band masking level Nm(i) . This is an approximation of the level of the curve of the masking level over the complete critical band (the real level of the masking level curve for a given signal can vary in a given critical band) . The masking level Nm(i) is defined according to the masking levels of the eight lower critical bands (Nm(i-8) to Nm(i-1)) if they exist, and the three upper bands (Nm(i+1) to Nm(i+3)), if they exist. We have Nm(i)=.SIGMA. Nm(j), where j positive integer index such that j .epsilon. [i-8, . . . , i-1, i+1, . . . , i+3], Nm(j)=10.sup.[Xnm(j)-Av(j)-Vf(j)]/20, Xnm(j)=20 log.sub.10 (Av(j))+5.69 dB (sound pressure), Av(j)=6.025+0.275*z(j) for the tonal lines, Av(j)=2.025+0.175*z(j) for the non-tonal lines, where Av(j) is the masking index of j.sup.th critical band and j and z(j) the ratio of the j.sup.th critical band, Vf(j)=(i-j-1)*(17-0.15*Xnm(j))+17, of j from i-8 to i-1, and Vf(i+1)=0.4*Xnm(i+18)+6, Vf(i+2)=17*Xnm(i+2)+6, Vf(i+3)=34*Xnm(i+3)+6. z(j) is a constant defined for each critical band and z(1)=0.62 dB, z(2)=1.8 dB, .sub.z(3)=2.4 dB, z(4)=3.6 dB, z(5)=4.7 dB, z(6)=5.8 dB, z(7)=6.7 dB, z(8)=7.7 dB, z(9)=8.9 dB, z(10)=10.0 dB, z(11)=10.9 dB, z(12)=12.0 dB, z(13)=13.1 dB, z(14)=14.0 dB, z(15)=14.9 dB, z(16)=15.8 dB, z(17)=16.7 dB, z(18)=17.7 dB, z(19)=18.8 dB, z(20)=19.8 dB, z(21)=20.9 dB, z(22)=22.2 dB, z(23)=23.2 dB and z(24)=23.9 dB. In general, the most masked critical bands are the high frequency bands of the audio-frequency spectrum which are masked by the low frequency bands, statistically more powerful. After this brief look at the auditory masking phenomenon and its modelling, an example of the implementation of the invention will now be described consisting of transmitting the data in the passband of a broadcast audio-frequency signal. The data can be either analog (musical patterns for example) or digital (that is binary data). The data may concern broadcast audio-frequency signals (for example the name of a radio station or the references of musical works transmitted by this station) and its purpose is to be perceived by the auditor, for example via a liquid crystal display. This data could also be service data for the signal broadcaster or the controlling authorities and be imperceptible to the listener. In the remainder of the description given as an example, it will be assumed that the data are binary data. These data will be relevant, for example, to the programmes broadcast by a radio station. A radio station generally transmits, in the direction of these listeners, audio-frequency signals modulated by conventional amplitude or frequency modulation techniques. These audio-frequency signals could be a song, a signature tune, the voice of a host, etc. The invention proposes to calculate, from the audio-frequency signal to be transmitted, for one or more critical bands Bi of the audio-frequency spectrum, the masking level or levels of this or these critical bands. If, for a critical band, the masking level is higher than the level of the audio-frequency signal, the corresponding part of the audio-frequency signal can be eliminated without a difference perceptible to the listener. The invention proposes to insert data (we shall speak of data audio-frequency signals) in a way inaudible to the listener into this critical band or a part of this critical band to replace the original audio-frequency signal (provided that, of course, the level of the audio-frequency signal of the data is lower than the critical band masking level). For the reception of the transmitted signal, it is sufficient to filter the signal received as a function of the critical bands to separate the data audio-frequency signal and process the transmitted data. We can see that the flow rate of the transmitted information cannot in practice be fixed, the original signal (and therefore the corresponding critical band masking levels Nm(i)) being a priori variable over time both in frequency and in amplitude. A data transmission system according to the invention will mainly include a data insertion device (an example of which is shown on FIG. 4) and the data reception device (an example of which is shown on FIG. 3). Typically, the data insertion device could be used either at the sound or visual broadcasting final control room stage or at the audio-frequency signal production stage. The data reception device will include for example a received data display device (if the data are intended for the listener) and/or a storage device (if the data are dedicated for example to a deferred audiometry control). The reception device could also include a device for retransmitting information, for example to a game remote for interactive television programmes. The data audio-frequency signal could be collected, at the reception device, either acoustically by a simple microphone (placed beside the loudspeaker of the radio receiver), or electrically using an appropriate connector (such as an audio recording output). In reference to FIG. 4, we will describe, as an example, a data insertion device 1, the information being in this case binary data. To transmit the data in the audio-frequency signal of a radio or TV programme, we replace, in certain frequency bands of this signal, the signal by a digital modulation. This transmission is preferably made at a level lower than the masking levels of these frequency bands in order to ensure the inaudible character of the transmitted information. Also, this transmission is preferably made when these masking levels are sufficiently high to ensure a satisfactory signal-to-noise ratio in relation to the broadcasting channel. In an example, the data to be transmitted could be organised into frames consisting of a start word and a defined number of data words. Also, a frame could be chosen including a start word, a variable number of data words, and an end word. The data insertion device 1 shown on FIG. 4 includes an input 2 to receive the original audio-frequency signal S to be transmitted (song, voice of a host, etc.), an input 3 to receive the data D to be transmitted, and an output 4 to deliver an audio-frequency output signal S' produced from the original audio-frequency signal S and the data D. The audio-frequency signal S is filtered by a bench of twelve bandpass filters FPB' 13 to FPB' 24, preferably complex, receiving at input the audio-frequency signal S. The analytical processing of the signal S facilitates the calculation of the amplitudes. Each complex filter produces at output the real part (R'.sub.13 to R'.sub.24) and the imaginary part (I'.sub.13 to I'.sub.24) of the audio-frequency signal S in the frequency band (called F'.sub.13 to F'.sub.24) that it lets through. As will be seen, the bank of complex bandpass filters FPB'.sub.13 to FPB'.sub.24 enables the components of the audio-frequency signal S in the frequency bands F'.sub.13 to F'.sub.24 to be eliminated to insert the data. These frequency bands (F'.sub.13 to F'.sub.24) are bands included in the critical bands B.sub.13 to B.sub.24. An amplitude calculation element OAC1 calculates the amplitudes A'.sub.j (j integer index from 13 to 24) from signals R'.sub.j and I'.sub.j delivered by the filters FPB'.sub.13 to FPB'.sub.24. The audio-frequency signal S is also filtered by a bank of twenty bandpass filters FPB.sub.5 to FPB.sub.24, preferably complex, receiving at input the audio-frequency signal S. Each complex filter produces at output the real part (R.sub.5 to R.sub.24) and the imaginary part (I.sub.13 to I.sub.24) of the audio-frequency signal S in the frequency band that it lets through. The bank of complex bandpass filters FPB.sub.5 to FPB.sub.24 enables the masking levels of the critical bands B.sub.13 to B.sub.24 to be calculated. This calculation is done from an amplitude calculation element OAC2 calculating the amplitudes A.sub.i (i integer index from 5 to 24) from signals R.sub.i and I.sub.i delivered by filters FPB.sub.5 to FPB.sub.24. These amplitudes are delivered to a calculating processor ON calculating the masking levels Nm(13) to Nm(24). The amplitudes A'.sub.13 to A'.sub.24 and the masking levels Nm(13) to Nm(24) are delivered to a control element OC which will compare them two at a time to determine if two amplitudes A'.sub.j1 and A'.sub.j2 exist lower than the corresponding masking levels Nm(j1) and Nm(j2) (j1 and j2 being two different integer indexes between 13 and 24). If this is the case, there is at least two frequency bands F'.sub.j1 and F'.sub.j2 in the audio-frequency spectrum for which signal S is inaudible. Signal S can then be filtered to eliminate these spectral components in these two frequency bands F'.sub.j1 and F'.sub.j2. To do this, the real components, called R'.sub.1 and R'.sub.2, are subtracted from the signal S in these two frequency bands F'.sub.j1 and F'.sub.j2 of the original signal S. These two components R'.sub.1 and R'.sub.2 are delivered via a multiplexing device MUXP receiving the components R'.sub.13 to R'.sub.24, each of these components being weighted so that all but two of them (R'.sub.j1 and R'.sub.j2) are cancelled. This MUXP device is controlled by the control element OC. These components (for example we have R'.sub.1 =R'.sub.j1 and R'.sub.2 =R'.sub.j2) are then subtracted from the signal S (this signal having been delayed to take into account the time required to pass through the filters and the multiplexing device) in two adders SM1 and SM2 so that an audio-frequency signal S'M=S-R'.sub.1 -R'.sub.2 is produced. This audio-frequency signal S'M is subjectively identical, for the listener perceiving it, to signal S. The assembly formed of the bandpass filters F'13 to F'24, the multiplexing device MUXP and the adders SM1 and SM2 acts as an adaptive band-stop filter vis-a-vis signal S. The frequency bands F'.sub.j1 and F'.sub.j2 being freed to allow insertion of data D, we will now look at this insertion. Conventionally, the binary data D will first of all be conditioned. Note that this conditioning operation is in any event independent of the freeing of the frequency bands F'.sub.j in the audio-frequency signal S. The data D to be transmitted are conditioned in a device MFB so as to be transmitted in required frame form (that is by inserting start and possibly end words, redundant codes, etc.). Then, two data audio-frequency signals S.sub.1 and S.sub.2 will be produced by means of a modulator MOD. The digital modulation used will be for example a QPSK (Quadrature Phase Shift Keying) modulation, the conditioned data, NRZ (Non Return to Zero) coded, modulating the phase of two frequency carriers included in bands F'.sub.j1 and F'.sub.j2, preferably corresponding to the centre frequencies of the used bands F'.sub.j1 and F'.sub.j2 (which allows the complete width of these bands to be used to transmit the data audio-frequency signals S.sub.1 and S.sub.2). This modulation step requires of course knowledge, via the control element OC, of the frequency bands freed in the spectrum of signal S. In parallel with the freeing of the bands F'.sub.j1 and F'.sub.j2, the masking levels Nm(13) and Nm(24) are delivered by the element ON to a multiplexing device MUXN which will produce two levels N'm=Nm(j1) and N"m=Nm(j2) at output. In order to take the modulation chosen to produce signals S.sub.1 and S.sub.2 into account, two coefficients N' and N" will be produced from coefficients N'm and N"m using an automatic gain control device CAG. By means of two multiplexers M1 and M2, two data audio-frequency signals S'.sub.1 =N'*S.sub.1 and S'.sub.2 =N"*S.sub.2 will then be produced. By summing signals S'.sub.1, S'.sub.2 and S'.sub.M in to adders SM3 and SM4, a signal S'=S-(R'.sub.1 +R'.sub.2)+(S'.sub.1 +S'.sub.2) is produced. The signal S' produced includes both the audible audio-frequency components of the original audio-frequency signal S and the data D (represented by S'.sub.1 and S'.sub.2) which are inaudible. Once signal S' has been produced, it will conventionally be modulated according to known techniques before being transmitted to the listeners' receivers. Note that as the gain applied to signals S.sub.1 and S.sub.2 is only proportional to the masking levels of the j.sub.1-th and j.sub.2-th bands F'.sub.j1 and F'.sub.j2, the amplitude level of signals S'.sub.1 and S'.sub.2 could be greater than the amplitude levels of the components of signal S which were removed. Preferably, bands F'.sub.13 to F'.sub.24 have same width to ensure a fixed transmitted data flow rate irrespective of the bands F'.sub.13 to F'.sub.24 used to transmit them. A same type of modulation can thus be used irrespective of the bands freed in signal S. In the example shown, the possibility of transmitting the data in the last twelve critical bands, from critical band B.sub.13 (fc=1860 Hz) to critical band B.sub.24 (fc=13750 Hz) is provided for. As has been seen, this information is transmitted in two bands each located in one of the twelve critical bands. Of course, the higher the number of bands F'.sub.j used simultaneously, the higher the transmitted data flow rate. A data insertion device using all freeable bands F'.sub.j can therefore be made. Nevertheless, it can be seen that simultaneous use of a reduced number of bands F'.sub.j enables the distortion probability of the original audio-frequency signal to be reduced if this signal varies to a high extent from one moment to another (although this probability is low on account of the temporal masking of the human ear). Whatever the critical band or bands in which the data are inserted, it is easy to understand that the band or bands F'.sub.j used within these critical bands have a width lower than or equal to the width of the corresponding critical bands. In the example shown, the first bench of bandpass filters consists preferably of bandpass filters F'.sub.13 to F'.sub.24 with bandwidths equal to 280 Hz at -3 decibels. This width corresponds to the width of the critical band usable to insert the data which has the lowest width, that is the width of the thirteenth critical band (of course, it is supposed here that the carrier frequencies used to produce the data audio-frequency signals are equal to the central frequencies of the critical bands). There is therefore little reason to use data transmission in the lower critical bands as their widths are lower and this would limit the maximum permissible flow rate. The bank of filters F'.sub.13 to F'.sub.24 is preferably achieved by multi-rate filtering thus giving a constant propagation time and a limited number of operations. The second bank of filters F.sub.5 to F.sub.24 is preferably obtained from reconstructible bandpass filters (that is filters such that the sum of the filtered output signals is the same as the input signal before filtering) the envelopes of which correspond to the critical bands. In other words, it is interesting to calculate as finely as possible the critical band masking levels to avoid producing data audio-frequency signals which could be audible. The binary information is for example grouped into words of thirty-two bits. A transmitted frame will include for example a start word, coded over thirty-two bits and a data word of thirty-two bits. The start word consists for example of the first nine bits comprising a lock-on ramp used in the reception device, the next twenty-three bits forming the synchronisation word. The data word consists for example of three bytes representing the data and a last redundancy byte for an error correction code if such a code is used. This organisation of the information frames corresponds to a transmission of information on the time frames of the audio-frequency signal lasting 256 milliseconds, which corresponds to the time required to transmit the sixty-four bits, that is two data frames. This enables a maximum binary flow rate of 500 bits per second to be attained. Preferably, the data frames are transmitted provided that the masking levels of the critical bands used to insert the data are greater than the minimum energy level providing resistance to the disturbances induced by the channel. Although not specified, it is of course preferable not to free frequency bands in the original audio-frequency signal when there is no data to be transmitted. For this, it is sufficient to cancel the signals produced at the output of the multiplexing device MUXP. Thus, even if the masking level of the original signal varies rapidly and extensively, there is no risk of disturbing the original signal by suppressing audible frequencies. Once the data transmission has been made, progressive cancellation of the output signals of the multiplexing device MUXP will preferably be carried out to reduce the probability to make the "filling in" audible. If the masking level of the original audio-frequency signal drops and the start word has been transmitted, transmission will be preferably continued to facilitate data processing at the level of the reception device. If the data are coded over thirty-two bits, this is not very troublesome on account of the temporal auditory masking. The data extraction device 5 shown on FIG. 3 includes an input 6 to receive the audio-frequency signal S'. The audio-frequency S' is filtered by a bank of twelve bandpass filters FPB".sub.13 to FPB".sub.24 with envelopes identical to the twelve filters FPB".sub.13 to FPB".sub.24. Twelve audio-frequency signals S'.sub.13 to S'.sub.24 are thus produced corresponding to the spectral components of signal S in bands F'.sub.13 to F'.sub.24 where it is likely that we shall find the data inserted by a device similar to the one described as reference on FIG. 4. Device 5 includes a bank of twelve demodulators DEMOD.sub.13 to DEMOD.sub.24, each demodulator being associated with one of the bandpass filters. Once the signals have been demodulated, they are sampled in samplers EC.sub.13 to EC.sub.24 associated with the clock recovery devices RC.sub.13 to RC.sub.24 to produce the binary data. Once the audio-frequency signals have been sampled, the binary data produced are processed in the recognition elements RTB.sub.13 to RTB.sub.24 to determine if these data are representative of the transmitted data (in which case the synchronisation bits of the start word will be present) or if these data correspond to nothing (the probability being fairly low that bits corresponding to the synchronisation bits of a start word can be produced by sampling from any audio-frequency signal). Of course, if the data transmitted are not digital data but analog data, such as a musical pattern for example, the data insertion and extraction devices will be adapted to suit. In particular, it will not be necessary to use the modulation, demodulation and sampling devices. These will be replaced by means for converting the data to be inserted into frequencies to adapt the frequencies of the data to the frequencies freed in the insertion device.
|
Same subclass Same class Consider this |
||||||||||