CN104361890A - Method for embedding and recognizing broadcast audio watermark - Google Patents

Abstract

The invention discloses a method for embedding and recognizing a broadcast audio watermark. The method includes determining a time range in which a watermark signal can be embedded according to time domain masking effect; determining amplitude of the watermark signal according to frequency domain masking effect. A microphone of wearable electronic equipment like a cellphone and a tablet computer which are portable equipment limited in operational capability and battery capacity can be used to receive voice, the watermark is restored through application software running on the wearable electronic equipment, and information can be extracted after the process of transmission with loss. The audio watermark is embedded in a frequency band audible for human ears and below 16000Hz according to the time domain masking effect and the frequency domain masking effect, in other words, the television media audio watermark technology the frequency band audible for the human ears and below 16000Hz is applied in a broadcast system. Carrier signal range for frequency spectrum modulation is expanded to be 4000-8000Hz.

Description

A kind of embedding of broadcast audio watermark and recognition methods

Technical field

The invention belongs to digital signal processing technique field, particularly relate to a kind of embedding and recognition methods of broadcast audio watermark.

Background technology

The sound signal frequencies scope of broadcast system is 0 to 16000 hertz, at this frequency band, the auditory system of normal person is very sensitive, if the people's ear audible frequency section below 16000 hertz embeds audio frequency watermark, the tonequality of carrier sound can be affected, this type of TV media audio watermarking technique is caused to be applied in broadcast system, so existing TV media audio watermarking technique embeds audio frequency watermark at the imperceptible frequency band of people's ear institute.In addition, existing broadcast audio digital watermark, as being applied to the portable audience rating measuring instrument (PPM of Ah 's wound market research agency, Portable People Meter) technology, need in test side to use specialized equipment, the portable equipment of arithmetic capability and limited battery capacity cannot be applied to, as in the wearable electronic such as mobile phone, panel computer.

More existing technology, usually watermark or vocal print are embedded in audio-video document, through Digital Transmission, as lossless manner such as data compression copy or Internet Transmissions, receiving end directly goes out watermark or vocal print from extracting data, and for the equipment of information extraction after the transmitting procedure by damaging, as the portable equipment such as mobile phone, panel computer using microphone to gather sound, embed not all roses with extractive technique.

Summary of the invention

The invention provides a kind of embedding and recognition methods of broadcast audio watermark, in the appreciable frequency band of people's ear, TV media audio watermarking technique cannot be applied in broadcast system to solve prior art, and also cannot be applied to the portable equipment of arithmetic capability and limited battery capacity, as the technical matters in the wearable electronic such as mobile phone, panel computer.

Technical scheme: a kind of embedding of broadcast audio watermark and recognition methods, comprise: determining can the time range of embed watermark signal: the average power calculating carrier voice signal in 4000 to 8000 frequency range, according to temporal masking, if the average power of described carrier voice signal is all greater than default threshold value within the predetermined time, then in the time range of 40 milliseconds, described carrier voice signal can be used for embed watermark signal; Determine the amplitude of described watermark signal: in the described time range embedding described watermark signal, calculate frequency domain masking threshold curve, calculate the amplitude of described watermark signal according to frequency domain masking threshold opisometer, described watermark signal is processed and obtains watermark signal to be embedded.

Further, embedding and the recognition methods of described a kind of broadcast audio watermark also comprise, obtain described watermark signal: carry out cyclic redundancy check (CRC) code coding and convolution coding by with the watermark information of binary representation, obtain the sequence of symhols of described watermark information, by the frequency range of described sequence of symhols spread-spectrum modulation to 4000 to 8000 hertz, obtain described watermark signal.

Further, embedding and the recognition methods of described a kind of broadcast audio watermark also comprise, obtain the time-frequency representation of described carrier voice signal: by described carrier voice signal in time domain point framing, every frame has 1024 sample points, the overlap of 50% is had between frame, frame by frame through short time discrete Fourier transform to frequency domain, obtain the time-frequency representation of described carrier voice signal.

Further, after determining to embed the time range of described watermark signal and the amplitude of described watermark signal, also comprise: on described carrier voice signal, superpose described watermark signal: in the described time range embedding described watermark signal, the watermark signal described to be embedded obtained after process is added on described carrier voice signal.

Further, embedding and the recognition methods of described a kind of broadcast audio watermark also comprise, the identification of watermark signal: be installed on the microphones voice signal of the application software on receiving end by receiving end, through short time discrete Fourier transform, spread-spectrum demodulation is carried out, to obtain described watermark information in the frequency range of 4000 to 8000 hertz.

Wherein, described receiving end carries out short time discrete Fourier transform to the voice signal received and the process of carrying out spread-spectrum demodulation comprises: at receiving end, to the described voice signal received, short time discrete Fourier transform is done with identical frame length and frame overlap, spread-spectrum demodulation is carried out the frequency ranges of 4000 to 8000 hertz, carry out Viterbi decoding and cyclic redundancy check (CRC), if verification makes mistakes, then illustrate that described watermark information does not exist; If verification succeeds, then obtain described watermark information.

Wherein, described predetermined time is 40 milliseconds, and described default threshold value is-40dB.

The beneficial effect that the present invention brings: the broadcast voice signal after embed watermark signal, transmit through broadcast system, received by radio again and play out, the portable equipment of available arithmetic capability and limited battery capacity, as the microphones sound of the wearable electronic such as mobile phone, panel computer, watermark is recovered by the application software operated in wearable electronic, can information extraction after the transmitting procedure damaged.And, with the same technology extracting watermark after damaging transmitting procedure, according to temporal masking and frequency domain masking effect, people's ear audible frequency section below 16000 hertz embeds audio frequency watermark, and TV media audio watermarking technique is applied in broadcast system by the people's ear audible frequency section namely below 16000 hertz.Audio frequency after embed watermark is analyzed, obvious watermark signal power is all can't see in any frequency range, the carrier signal scope of spread-spectrum modulation is 4000Hz-8000Hz, take frequency range few, easier embed watermark, not only not easily discovered by people's ear, and also not easily discovered concerning universal audio analysis tool, i.e. good confidentiality.

Accompanying drawing explanation

Fig. 1 is embedding and the recognition methods block diagram of a kind of broadcast audio watermark of the present invention;

Fig. 2 is embedding and the recognition methods process flow diagram of a kind of broadcast audio watermark of the present invention.

Embodiment

The following description and drawings illustrate specific embodiment of the invention scheme fully, to enable those skilled in the art to put into practice them.Embodiment only represents possible change.Unless explicitly requested, otherwise independent parts and function are optional, and the order of operation can change.The part of some embodiments and feature can be included in or replace part and the feature of other embodiments.

For a better understanding of the present invention, brief description masking effect, masking effect refers to that the ear of people is only responsive to the most obvious audio response, and for insensitive sound, reaction is not more then responsive.Such as in the whole frequency spectrum of sound, if the sound of some frequency bands is stronger, then people is just insensitive to the sound of other frequency band.The weak pure tone of sounding simultaneously in its vicinity can be sheltered in a strong pure tone, this characteristic is called that frequency domain is sheltered, except having except occlusion between the sound that sends simultaneously, also occlusion is had between sound adjacent in time, be called time domain masking, producing the main cause of time domain masking is that the brain process information of people needs to spend the regular hour.

In some illustrative embodiments, as shown in Fig. 1 to 2, wherein, Fig. 1 illustrates the flow process that embodiment is concrete in form of a block diagram, and Fig. 2 shows the flow process of embodiment entirety in flow diagram form, provides a kind of embedding and recognition methods of broadcast audio watermark, comprising:

101: obtain watermark signal.Watermark information, with binary representation, carries out cyclic redundancy check (CRC) code coding 1011 and convolution coding 1012, obtains the sequence of symhols w of described watermark information _k, wherein, described cyclic redundancy check (CRC) code coding and convolution coding belong to the error control method of digital communication, are conducive to receiving end and find or correct the error code that may exist in the letter code received.Described sequence of symhols w _kagain through the frequency range 1013 of spread-spectrum modulation to 4000 to 8000 hertz, obtain described watermark signal w _k(t).

Wherein spread-spectrum modulation detailed process is as follows:

$w_k\left(t\right)=\underset{k=K1}{\overset{K2}{\mathrm{\Σ}}}{w}_k\×C_k\left(t\right)$

Wherein, C _kt () is the carrier wave of spread-spectrum modulation, K1 and K2 is determined by sampling rate, and computation process is as follows:

Wherein, int represents round numbers part.

102: the time-frequency representation obtaining carrier voice signal.Carrier voice signal represents with x (t), divide framing by described carrier voice signal x (t) in time domain, every frame has 1024 sample points, has the overlap of 50% between frame, frame by frame through short time discrete Fourier transform to frequency domain 1021, obtain the time-frequency representation X of carrier voice signal _tk (), wherein t is time index, and k is frequency index.

Wherein short time discrete Fourier transform is as follows:

$X_t\left(k\right)=\frac{1}{1024}\underset{n=0}{\overset{1023}{\mathrm{\Σ}}}{x}_t\left(n\right)e^{-j\frac{2\mathrm{\πnk}}{1024}},k=\mathrm{0,1},...,1023$

103: determine the time range that can embed described watermark signal.The average power EX of carrier voice signal in 4000 to 8000 frequency range is calculated by the time-frequency representation of described carrier voice signal _t1031, according to temporal masking find described can the time range 1032 of embed watermark signal.Concrete grammar is: according to temporal masking, if the average power EX of described carrier voice signal _tall be greater than default threshold value within the predetermined time, then, after in time of 40 milliseconds, described carrier voice signal can be used for embedding described watermark signal.Wherein, described predetermined time is 40 milliseconds, and described default threshold value is-40dB, if time starting point is designated as T _w, unit is millisecond, at T _w+40and T _w+80can insert described watermark signal in the time range of millisecond, the duration is 40 milliseconds.

Wherein calculate the average power EX of carrier voice signal _tdetailed process as follows:

${\mathrm{EX}}_t=\frac{1}{K2-K1+1}\×\underset{k=K1}{\overset{K2}{\mathrm{\Σ}}}20\×\log \left(\right|X_t\left(k\right)\left|\right)$

104: the amplitude determining described watermark signal.In order to make the described watermark signal being embedded into described carrier voice signal be not the perception of people's ear institute, the amplitude A W of described watermark signal must be adjusted _k, make it be sheltered by described carrier voice signal.Method is: according to frequency domain masking effect, in the described scope embedding described watermark signal, namely at described T _w+40and T _w+80frequency domain masking threshold curve M is calculated in the time range of millisecond _t(k) 1041.According to frequency domain masking threshold curve M _tk () calculates the amplitude A W of described watermark signal _k1042.Concrete computation process is as follows:

With reference to the psychoacoustic model of MPEG-1Part3 (ISO/IEC11172-3), at T _w+40and T _w+80by the time-frequency representation X of described carrier voice signal in the time range of millisecond _tk () calculates frequency domain masking curve M _t(k).

According to frequency domain masking threshold curve M _tk () calculates the amplitude A W of described watermark signal _kprocess as follows:

ratio_w_x _t(k)＝MX _t(k)-20log(|X _t(k)|)

$A{W}_k=X_t\left(k\right)\×{10}^{\mathrm{ratio}\_w\_x_t\left(k\right)/20}$

With the amplitude A W of described watermark signal _kbe multiplied by described watermark signal w _kt (), obtains watermark signal 1043 to be embedded.

105: on described carrier voice signal, superpose described watermark signal.In the time range that can embed described watermark signal, i.e. described T _w+40and T _w+80in the time range of millisecond the watermark signal to be embedded after described process to be added on described carrier voice signal 1051, to obtain carrier voice signal y (t) containing described watermark signal.

Wherein, described carrier voice signal superposes the computation process of described watermark signal to be embedded, as follows:

y(t)＝x(t)+AW _k×w _k(t)

Finally by the described carrier voice signal containing watermark signal, through broadcast system transmission, receive with radio and play back.

106: the identification of described watermark signal.At receiving end, be installed on microphones voice signal r (t) of the application software on receiving end by receiving end, do short time discrete Fourier transform 1061 with identical frame length and frame overlap, carry out spread-spectrum demodulation 1062 the frequency ranges of 4000 to 8000 hertz.Carry out Viterbi decoding and cyclic redundancy check (CRC) 1063 simultaneously, if verification makes mistakes, illustrate that described watermark information does not exist, if verification succeeds, namely obtain described watermark information.Wherein, described Viterbi decoding is a kind of method to convolutional code decoder.

Wherein short time discrete Fourier transform is as follows:

$R_t\left(k\right)=\underset{n=0}{\overset{1023}{\mathrm{\Σ}}}{r}_t\left(n\right)e^{-j\frac{2\mathrm{\πnk}}{1024}},k=\mathrm{0,1},...,1023$

Wherein spread-spectrum demodulation is as follows:

$r_k=\underset{k=K1}{\overset{K2}{\mathrm{\Σ}}}{R}_t\left(k\right)\×C_k\left(t\right)$

In some illustrative embodiments, described receiving end is mobile phone or panel computer.Watermark is recovered by the application software operated in wearable electronic, can information extraction after the transmitting procedure damaged, in test side without the need to using specialized equipment, broadcast audio digital watermark is made to be applied to the portable equipment of arithmetic capability and limited battery capacity, as in the wearable electronic such as mobile phone, panel computer.

Above-described embodiment is the present invention's preferably embodiment; but embodiments of the present invention are not restricted to the described embodiments; other are any do not deviate from Spirit Essence of the present invention and principle under the change made; modify; substitute, combination, simplify; all should be the substitute mode of equivalence, all should be included in protection scope of the present invention.

Claims (8)

1. the embedding of broadcast audio watermark and a recognition methods, is characterized in that, comprising:

Determining can the time range of embed watermark signal: the average power calculating carrier voice signal in 4000 to 8000 frequency range, according to temporal masking, if the average power of described carrier voice signal is all greater than default threshold value within the predetermined time, then in the time range of 40 milliseconds, described carrier voice signal can be used for embed watermark signal;

Determine the amplitude of described watermark signal: in the described time range embedding described watermark signal, calculate frequency domain masking threshold curve, calculate the amplitude of described watermark signal according to frequency domain masking threshold opisometer, described watermark signal is processed and obtains watermark signal to be embedded.

2. the embedding of a kind of broadcast audio watermark according to claim 1 and recognition methods, is characterized in that, also comprise:

Obtain described watermark signal: carry out cyclic redundancy check (CRC) code coding and convolution coding by with the watermark information of binary representation, obtain the sequence of symhols of described watermark information, by the frequency range of described sequence of symhols spread-spectrum modulation to 4000 to 8000 hertz, obtain described watermark signal.

3. the embedding of a kind of broadcast audio watermark according to claim 1 and recognition methods, is characterized in that, also comprise:

Obtain the time-frequency representation of described carrier voice signal: by described carrier voice signal in time domain point framing, every frame has 1024 sample points, have the overlap of 50% between frame, frame by frame through short time discrete Fourier transform to frequency domain, obtain the time-frequency representation of described carrier voice signal.

4. the embedding of a kind of broadcast audio watermark according to claim 2 and recognition methods, is characterized in that, after determining to embed the time range of described watermark signal and the amplitude of described watermark signal, also comprise:

Described carrier voice signal superposes described watermark signal: in the described time range embedding described watermark signal, the watermark signal described to be embedded obtained after process is added on described carrier voice signal.

5. the embedding of a kind of broadcast audio watermark according to claim 4 and recognition methods, is characterized in that, also comprise:

The identification of watermark signal: be installed on the microphones voice signal of the application software on receiving end by receiving end, through short time discrete Fourier transform, carry out spread-spectrum demodulation in the frequency range of 4000 to 8000 hertz, to obtain described watermark information.

6. the embedding of a kind of broadcast audio watermark according to claim 5 and recognition methods, is characterized in that, described receiving end carries out short time discrete Fourier transform to the voice signal received and the process of carrying out spread-spectrum demodulation comprises:

At receiving end, to the described voice signal received, do short time discrete Fourier transform with identical frame length and frame overlap, spread-spectrum demodulation is carried out the frequency ranges of 4000 to 8000 hertz, carry out Viterbi decoding and cyclic redundancy check (CRC), if verification makes mistakes, then illustrate that described watermark information does not exist; If verification succeeds, then obtain described watermark information.

7. the embedding of a kind of broadcast audio watermark according to claim 5 or 6 and recognition methods, is characterized in that, described receiving end is mobile phone or panel computer.

8. the embedding of a kind of broadcast audio watermark according to claim 1 and recognition methods, it is characterized in that, described predetermined time is 40 milliseconds, and described default threshold value is-40dB.