CALIKUS, Onur (Turkcell Iletisim Hizmetleri Anonim Sirketi, Turkcell PlazaMesrutiyet Caddesi No:7, Tepebasi Istanbul, 34755, TR)
SAHIN, Coskun (Turkcell Iletisim Hizmetleri Anonim Sirketi, Turkcell PlazaMesrutiyet Caddesi No:7, Tepebasi Istanbul, 34755, TR)
KURT, Gunes (Turkcell Iletisim Hizmetleri Anonim Sirketi, Turkcell PlazaMesrutiyet Caddesi No:7, Tepebasi Istanbul, 34430, TR)
CALIKUS, Onur (Turkcell Iletisim Hizmetleri Anonim Sirketi, Turkcell PlazaMesrutiyet Caddesi No:7, Tepebasi Istanbul, 34755, TR)
SAHIN, Coskun (Turkcell Iletisim Hizmetleri Anonim Sirketi, Turkcell PlazaMesrutiyet Caddesi No:7, Tepebasi Istanbul, 34755, TR)
| CLAIMS A watermark embedding method characterized in that information signal (B) is embedded to the high energy parts of the host signal (A) that will be watermarked, such that the signal is framed in the time and frequency dimension, in a multi-carrier system comprising a transmitter (1) having at least one energy analysis block (1 1 ), at least one modulator (12), at least one error control coder (13), a combiner (14), and depending on the application, at least one information generator (15) that generates information; and a receiver (2) having at least one energy detector (21) which detects beginning and end of the signal, at least one frequency analysis block (22), at least one filter (23), at least one demodulator (24), a data combiner (25), an error control decoder (26), and depending on the application, at least one information interpreter (27) that reconstructs the incoming information (B). A system used in the method according to Claim 1 and characterized by a demodulator (24) comprising at least one coarse acquisition unit (241), at least one matched filter (242), at least one sampler (243), at least one threshold estimation unit (244), at least one fine acquisition unit (245), at least one decision unit (246), at least one equalizer (247), and at least one interference canceller (248). A system according to Claim 2, characterized by an energy analysis block (1 1); which determines the high energy parts of the audio file which is the host signal (A) that will be used in watermarking and the time-frequency frames that may be used according to parameters such as window, amplitude, waveform; and transmits the concerned parameter information of the frames it has determined to the modulator (12) for watermarking. A system according to Claim 3, characterized by a multi-carrier modulator (12) which uses features of the audio file (A) to generate the data to be adaptively embedded into the audio file (A). A system according to Claim 4, characterized by a combiner (14) which places the multi-carrier watermark generated by the modulator (12) to the high energy parts of the host signal (A) to be used, such that the watermark is framed in the time and frequency dimension. A system according to Claim 5, characterized by a modulator (12) which computes the watermark signal according to the equation of d [k] =∑=1∑„ b naiWi [fc]cos(27r/;-fc Fs) p( (fc - nTj A system according to Claim 6, characterized by a combiner (14), which, after the watermark signal is generated in the modulator (12), receives the signal from the modulator, embeds the watermark signal (d [k] ) to the host signal(SM ) and sends it to the speaker (H). A system according to Claims 2 to 7, characterized by an energy detector (21), which determines the energy levels of the watermarked signal received by the receiver sensor (M), locates the silent parts that occur in the beginning and end of the watermark signal by checking the energy of the signal, and removes them from this recorded signal. A system according to Claim 8, characterized by a data combiner (25) which enables to recombine the data (C)I ...L that are divided into L carriers and time frames. A system according to Claim 9, characterized by a coarse acquisition unit (241) which approximately determines the time windows in the signal which is filtered in accordance with specific frequencies and such that it will not contain silent parts. A system according to Claim 10, characterized by a fine acquisition unit (245) which detects the time windows in the signal, which does not include silent parts and which is filtered in accordance with specific frequencies, in intervals smaller than Tt time. 12. A system according to Claim 1 1 , characterized by a receiver (12) which computes the watermarked signal according to the equation of r[k] = t[k] * h[k] + n[k] |
Field of the Invention The present invention relates to a method for embedding information into audio signals in multi-carrier systems.
Prior Art There are various systems for embedding watermarks into audio signals (audio watermarking). In some of these systems, a plurality of carriers is used to generate the watermark. One of these systems is disclosed in the United States patent document no. US6915002B2. This document discloses a watermark embedder which generates the watermark by encoding through more than two channels and embeds the said watermark into the host signal.
Another system which embeds information into audio signals is defined in the United States patent document no. US2004204943. There is provided an energy level detector in the system described in this document. With this detector, energy of the signal is measured in the given time domain and it is decided whether or not watermarking will be performed. When extremely different energy levels are detected in the given time domain, watermarking is not performed in order to avoid audible time dispersion. In the state of the art applications, high computational complexity of the receiver that should be employed in the system impedes use of this receiver in mobile devices (e.g. cellular phones) with low capabilities or limited energy use.
In addition to this, the receiver performance may be low due to the interference dependant on the audio signal (A) used in the transmitter (Figure 1 and 2). Summary of the Invention
The objective of the present invention is to realize a watermark embedding method which marks the audio by using multi-carriers in receivers and transmitters.
Another objective of the present invention is to realize a watermark embedding method wherein the audio signal is framed in time and frequency dimension and an information signal is embedded into the marked region.
A further objective of the present invention is to realize a watermark embedding method which uses a receiver with a simple structure that extracts the information from the audio by using the frames in the time and frequency dimension. Another objective of the present invention is to realize a watermark embedding method wherein the performance loss due to the interference occurring in the receiver is prevented.
Detailed Description of the Invention
The watermark embedding method realized to fulfill the objective of the present invention is illustrated in the accompanying figures wherein,
Figure 1 is the schematic view of the transmitter (Γ) comprising an encoder (K) in the state of the art.
Figure 2 is the schematic view of the receiver (2') comprising a demodulator in the state of the art.
Figure 3 is the schematic view of the transmitter used in the watermark embedding method of the present invention.
Figure 4 is the schematic view of the receiver used in the watermark embedding method of the present invention. Figure 5 is the schematic block diagram of the demodulator of the receiver used in the watermark embedding method of the present invention.
Figure 6 is the energy analysis graphic of a sample audio file into which a watermark will be embedded by means of the watermark embedding method of the present invention.
Figure 7 is an energy analysis graphic which illustrates, with frames, the positions of the watermarks to be embedded into the sample audio file of Figure 6 by means of the watermark embedding method of the present invention. The parts illustrated in the figures are each given a reference numeral where the numerals refer to the following:
1. Transmitter
11. Energy analysis block
12. Modulator
13. Error Control Coder
14. Combiner
15. Information generator
2. Receiver
21. Energy detector
22. Frequency analysis block
23. Filter
24. Demodulator
241. Coarse acquisition unit
242. Matched Filter
243. Sampler
244. Threshold estimation unit
245. Fine acquisition unit
246. Decision unit
247. Equalizer
248. Interference canceller 25. Data Combiner
26. Error control decoder
27. Information interpreter
A: Host signal
B: Information signal
C: Reconstructed information signal
D: Action
H: Source of Sound (Speaker)
M: Receiver sensor
(a) L: The host signal portion carried by each carrier (there are "L" number of "a" signal, where the number "L" is a natural number equal to or greater than one and refers to the number of carriers)
(b) Li The information signal portion carried by each carrier (there are "L" number of "b" signal, where the number "L" is a natural number equal to or greater than one and refers to the number of carriers)
(C) L: The reconstructed information signal portion in each demodulator (there are "L" number of "c" signal, where the number "L" is a natural number equal to or greater than one)
(a+b)fL : The signal portion exiting from each filter (there are "L" number of ""(a+b)fL" signal, where the number "L" is a natural number equal to or greater than one and refers to the number of filters)
In the method of the present invention, a transmitter (1) and a receiver (2) are used. The transmitter (1) comprises at least one energy analysis block (1 1), at least one modulator (12), at least one error control coder (13), a combiner (14), and depending on the application, at least one information generator (15) that generates information.
The receiver (2) comprises at least one energy detector (21) which detects beginning and end of the signal, at least one frequency analysis block (22), at least one filter (23), at least one demodulator (24), a data combiner (25), an error control decoder (26), and depending on the application, at least one information interpreter (27) that reconstructs the incoming information (B). An action (D) is performed in accordance with the data formed in the interpreter (27). The action (D) is routed to the user depending on the data formed. The receiver (2) is configured to be used in a mobile device such as a cellular phone. For example, if the data formed in the interpreter (27) is a "00" value of two bits, prepaid minutes are loaded to a GSM subscriber and this is reported to the mobile device of the concerned subscriber via a message. As another example, if the data formed in the interpreter (27) is a "1 1" value of two bits, a value added service subscription is given to the GSM subscriber and this is reported to the mobile device of the concerned subscriber via a message.
The demodulator (24) comprises at least one coarse acquisition unit (241), at least one matched filter (242), at least one sampler (243), at least one threshold estimation unit (244), at least one fine acquisition unit (245), at least one decision unit (246), at least one equalizer (247), and at least one interference canceller (248).
The energy analysis block (1 1) in the transmitter (1) determines the high energy parts of the audio file which is the host signal (A) that will be used in watermarking and the time-frequency frames that may be used. The energy analysis block (1 1) determines these frames according to parameters such as window, amplitude, waveform. The said block (11) transmits the concerned parameter information of the frames it has determined to the modulator (12) for watermarking.
The information generator (15) is the unit which generates the information (B), i.e. bit values, pertaining to the watermark that is intended to be sent. The said unit generates different information depending on the application.
The error control coder (13) attenuates the impact of the errors occurring in the information (B) bits depending on the impact of the channel and the current noise level.
The modulator (12) is a multi-carrier modulator which uses features of the audio file (A) to generate the watermark to be adaptively embedded into the audio file (A).
The combiner (14) places the multi-carrier watermark generated by the modulator (12) to the high energy parts of the host signal (A) to be used, such that the watermark is framed in the time and frequency dimension. The watermark signal is obtained in the modulator (12) by using the following equation (1) in the preferred embodiment of the invention: The variables in equation (1) are defined below:
d watermark signal
number of carriers that are used
ki.r. , bit of the information signal (B) belonging to carrier i a i, amplitude of the carrier i
.[k], window function of the carrier i ft carrier frequency value of the carrier /
sampling frequency (generally 22050 Hz value is used) waveform used by the carrier i in the transmitter (1) carrier wavelength of the carrier i (period)
n Symbol number
After the watermark signal is generated in the modulator (12), the watermark signal is transmitted from the modulator (12) to the combiner (14). Watermark signal (d [k] ) is embedded into the host signal ( S W ) in the combiner (14). Then the combiner (14) transmits the host signal ( S M ) and the watermark signal (d [k] ) that it has combined to the speaker (H) for the purpose of being broadcasted.
The signal to be transmitted from the speaker (H) is given in the following equation (2).
t[k] = s [k] + d[k] (2)
The variables in equation (2) are defined below:
s [k] signal transmitted from the speaker (H)
d [k] watermark signal
5 [k] host signal
In one example of application of the invention, chicken sound is used as the host signal (A). Figure 6 shows the distribution of energy values of the chicken sound file by time and frequency. In this sound file, two carriers are hidden next to the high energy sound components (Figure 7). Parameters of the said hidden data (information) are shown in Table 1.
Table 1 : Time and frequency positions of example watermark parameters embedded into the chicken sound (Figure 6) chosen as the host signal (A) Fc (Hertz) (Carrier Frequency) Time Frame (second)
950 2.494 - 2.766
1322 2.63- 3.174
Energy levels of the watermarked signal received by the receiver sensor (M) are determined in the energy detector (21) in the receiver (2). Additionally, upon recording of the entire watermarked signal, there might occur silent parts in the beginning and end thereof. These silent parts are located by checking the energy of the signal in the energy detector (21) and are removed from this recorded watermarked signal.
Although the sent information signal (B) is in a specific time/frequency frame, there might be a deviation in the received frequency values due to the hardware in the receiver (2) and the transmitter (1). The frequency analysis block (22) enables to detect these deviations and to obtain the correct frequency value.
The filter (23) allows passage of the signals (a+b)n...fL that are correct in accordance with the frequency values in the frequency analysis block (22), while not allowing passage of the other signals.
The data combiner (25) is the unit which enables to recombine the data (C)i. that are divided into multi-carriers (L number) and time frames.
The error control decoder (26) attenuates the impact of the errors occurring in the bits depending on the impact of the channel and the current noise level.
The information interpreter (27) is the unit which interprets the bit values received from the error control decoder (26) and reconstructs the watermark signal (C). The coarse acquisition unit (241) located in the demodulator (24) is the unit which approximately determines the time windows in the signal which is filtered in accordance with specific frequencies and such that it will not contain silent parts. The signal framed in time and frequency dimension is sent from the coarse acquisition unit (241) to the matched filter (242).
The matched filter (242) is used to obtain the ^ί,π values of the information (B). The data leaving the matched filter (242) is sent to the sampler (243). In this unit, data are sampled in certain time intervals, for example in T-i or shorter time intervals.
The threshold estimation unit (244) adaptively decides on the values of parameters such as window, amplitude, waveform that are used for correct operation of the receiver (2).
The fine acquisition unit (245) is the structure which detects the time windows in the signal, which does not include silent parts and which is filtered in accordance with specific frequencies, in intervals smaller than T i time.
The decision unit (246) decides on the values that correspond, in the system, to the bi,n values sent from the matched filter (242) in accordance with the samples generated by the sampler (243). For example, in a binary system where +1, -1 is sent, when the matched filter (242) generates a value of 0,8, the decision unit (246) decides which value it will be used as. For instance, according to the system to be used, since 0,8 is closer to +1 , the unit may decide that 1 is sent. Or, if a value closer to -1 is generated by the matched filter (242), the decision unit (246) may decide that 0 is sent. The equalizer (247) is the unit which attenuates or eradicates the distorting effects of the multipath channel on the estimated information signal. The interference canceller (248) is the unit which produces the effects of A that has passed through the radio communication channel from the received filtered (A+B) signal and subtracts its effect from the signal received over the air by means of the receiver sensor ( ), whereby decreases error rate and enhances system performance. By means of the said unit (248), the audio is enabled to be saved from effects like power loss, echoes that occur during transmission over the air until the audio arrives from the speaker (H) to the microphone (M). The watermarked signal received by the receiver sensor (M) in the receiver is transmitted to the energy detector (21).
In one embodiment of the invention, the watermarked signal is computed in the receiver (2) as in equation (3) given below. r[k] = t-lk] * h[k] + n[k] (3) r[k] Watermarked signal received by a receiver sensor (M), for example by a microphone, and sampled digitally.
channel impulse response
* Convolution operator
n[k] Additive White Gaussian Noise value can be shown as follows by using equation (2) r[k] = d[k] * h[k] + s[k] * h[k] + n[k] = d[k]* h[k] + Interference + Noise
(4) Assuming that the terms interference and noise specified in equation (4) do not impede reception of the signal, the data 1 * .« J sent by the help of the demodulator (24) can be obtained in the receiver (2). It is also possible to use a different structure as the demodulator (24) in the invention. The demodulator (24) given here, is used as an example due to the operational simplicity of its structure (Figure 5). s[k] * h[k] j s d e f ine( j as the interference term.. After channel estimation is performed in the equalizer (247), this estimation h[k] is transmitted to the interference canceller (248). In the case that the samples of the audio signal planned to be used, which are framed in the time and frequency dimension, s[fe] * i[&] t er m may be generated in the interference canceller (247). After obtaining the value of the s lk] * h[k] te rm, use G f r[k] - s[fc] * h[k] va i ue i n the receiver will enhance system performance.
In the transmitter (1) used in the method of the present invention, although energy level of the embedded watermark signal is higher than the energy level of the host signal (A) (original signal), the perceived sound does not cause any disturbance.
It is possible to develop a wide variety of embodiments of the inventive method. The invention cannot be limited to the examples described herein. The invention is essentially according to the claims.