CN105721090B - A kind of detection and recognition methods of illegal f-m broadcast station - Google Patents

Abstract

The invention discloses a kind of detection and recognition methods of illegal f-m broadcast station, the present invention utilizes Digital Signal Processing and the information processing technology, exploitation to design a detection and recognition methods steady, realized convenient for software or hardware design.Utilize the characteristic of cycle repetition of illegal f-m broadcast station, based on related progressive algorithm, detect number of repetition and the time point of illegal f-m broadcast station, to complete the detection and identification of illegal f-m broadcast station, the accuracy of detection and identification is high, and it handles quickly, there is substantive distinguishing features outstanding and significant progress.

Description

A detection and identification method for illegal FM radio stations

technical field

The present invention relates to the technical field of radio monitoring in a wireless communication system, and relates to a method for detecting and identifying illegal FM radio stations, which realizes the automatic identification of illegal FM radio stations (commonly known as black broadcasting), and provides reasonable information for radio monitoring management and frequency spectrum resources. Provide technical support effectively.

Background technique

Illegal FM radio stations, commonly known as "black broadcasts", are mostly set up for illegal economic benefits, and the vast majority of broadcast content is false medical advertisements. From the perspective of social impact, black broadcast content will mislead the audience (consumers), and some black broadcast content is vulgar and seriously pollutes the social atmosphere. From the perspective of the order supervision of electromagnetic waves in the air, since the equipment of illegal FM radio stations are all "three noes" products, and the transmission power is extremely large (the transmission power of black radio stations is mostly above 1 kilowatt), it is easy to interfere with normal civil aviation, broadcasting, etc. interference with radiocommunication services. A prominent typical case is: the civil aviation frequency band (108MHz-117.975MHz aeronautical radio navigation service 117.975MHz-137MHz aviation mobile service) is adjacent to the FM broadcast frequency band (87MHz-108MHz), and illegal FM broadcast signals may interfere with the civil aviation frequency band and affect civil aviation airliners The navigation and command and dispatch of the aircraft will endanger the safety of the aircraft's take-off and landing. Therefore, the social harm and security risks of illegal broadcasting stations are extremely great.

At present, the detection and identification of illegal FM radio stations all adopt the method of manual identification, and judge whether the radio station is an illegal station by manually monitoring the broadcast content of the radio station for a certain period of time. This method is time-consuming and labor-intensive, and the efficiency is very low. Due to the rapid development of current electronic information equipment and technology, the cost of setting up an illegal FM radio station is getting lower and lower, and the difficulty is getting smaller and smaller. Based on the temptation of huge economic interests, the number of illegal FM radio stations in all large and medium-sized Cities have shown a rapid growth trend in recent years. This brings more difficulties and pressures to the monitoring and identification of illegal FM radio stations.

Contents of the invention

The purpose of the present invention is to provide a method for detecting and identifying illegal FM broadcasting stations, so as to solve the problem of inefficiency and deficiency in manual detection and identification of illegal broadcasting stations. The invention utilizes digital signal processing and information processing technology to develop and design a robust detection and identification method that is convenient for software or hardware design and realization.

In order to achieve the above object, the technical scheme adopted in the present invention is as follows:

A method for detecting and identifying illegal FM radio stations, comprising the following steps:

(1) sample data collection, carry out continuous voice recording to the target broadcasting station, and store as sample data files;

(2) sample data preprocessing, the sample data file is divided into multiple data sub-blocks, and each data sub-block is numbered and stored;

(3) Determine the reference data segment, and obtain the voice segment data of the starting time X of each data sub-block in turn, as the reference voice segment;

(4) Correlation waveform is obtained and calculated, and each reference speech segment is correlated with the data of all data sub-blocks point by point, and the correlation waveform signal of each reference speech segment is obtained;

(5) The detection of the correlation peak, for the correlation waveform signal of each reference speech segment, find and obtain the maximum value, and find the local maximum peak value whose amplitude value in the correlation waveform signal exceeds the maximum value * preset correlation peak threshold value point, and record its position, the number of local maximum peak points is the number of repetitions of the reference speech segment in the entire sample data. The position of the peak point of the local maximum is the position of the repeated voice segment that is basically identical to the reference voice segment in the entire recording period, and these position information are stored and recorded as relevant detection records;

(6) The number of times in the repeated recording information results of all reference speech segments is accumulated and summed to obtain the number of repetitions value, and then the number of repetitions value is compared with the preset number of repetitions threshold value to judge whether the target broadcasting station is is an illegal broadcasting station;

(7) Repeat steps (1)-(6) for other target broadcasting stations to obtain the judgment results of all target broadcasting stations. Thereby, the illegal radio stations with great suspicion are screened out automatically. Through the relevant detection and recording information of each radio station, it is possible to perform selective playback and playback of voice clips for repeated positions for further verification.

The invention utilizes the periodic repetition characteristic of the illegal FM broadcasting station and detects the repetition times and time points of the illegal FM broadcasting station based on a correlation accumulation algorithm, thereby completing the detection and identification of the illegal FM broadcasting station.

Specifically, in the step (1), the time for continuous voice recording is 3-5 days, the sampling rate of sample data is 22050 Hz, and the bit width of data samples is 16 bits.

Further, in the step (1), the sample data collection process of the target broadcasting station is specifically: first, receiving the FM broadcast signal of the target broadcasting station, and then performing digital down-conversion and decimation filtering on the FM broadcast signal, reducing the data rate It is 200KHz, and then FM demodulation processing is performed to obtain the original broadcast voice signal, and then low-pass filtering and variable rate processing are performed, the data rate is reduced to 22050Hz, and the data sample bit width is 16bit.

Still further, in the step (2), dividing the sample data file into multiple data sub-blocks is divided in units of hours. Each data sub-block divided in the step (2) is also subjected to data sample point extraction, and the extracted data sample points are indexed and named according to the number of the data sub-block, and stored as a new data sub-block. The data sub-blocks for subsequent correlation calculations and the reference voice data are both extracted data sub-blocks. Without affecting the relevant detection results, a certain degree of extraction of the original sample data can effectively reduce the amount of calculation and save calculation time, thereby achieving faster detection and identification.

Further still, in described step (3), at first, calculate the data length of the voice segment of time X according to sampling rate, data sample bit width, extraction rate, then obtain the start of each data sub-block according to the data length Speech segment data at time X. It should be noted that since the data sub-block is extracted data, the reference speech segment is also extracted data.

Still further, the processing object of the correlation operation, whether it is all the data sub-block data or the reference speech segment data based on the start time length X of the data sub-block data, is the extracted data. In the step (4), the fast Fourier transform calculation is used for the correlation calculation, which can reduce the number of multiplication operations to one percent or even one thousandth of the original, thereby greatly reducing the calculation amount of the correlation operation. The specific calculation process is:

(401) At first select the reference speech segment ref that needs to carry out correlation calculation, there are N data sub-blocks altogether, also just have N reference speech segments, n≤N, make n be initial value 1;

(402) Obtain all the data of the data sub-block n, obtain the voice segment data of the start time X of the data sub-block n+1, splice the two, and obtain the sample data b_n that needs to be compared with the reference voice segment;

(403) Carry out correlation calculation to reference speech segment ref and b_n, obtain correlation calculation result corr_bn;

(404) add n to 1, and judge whether n is equal to the total number of data sub-blocks N, if less than N, then return to step (402); if equal to N, then end, and all related calculation results corr_bn, n=1 , 2, 3, ... N, spliced into a complete data, this data is the correlation calculation result of the reference speech segment ref relative to the entire sample data - the correlation waveform signal corr.

In addition, in the step (403), the correlation calculation is based on the fast Fourier transform calculation, and the processing objects of the fast Fourier transform are all extracted data, and the specific process of the correlation calculation is:

(411) Acquire the processing object, i.e. the voice segment data of the start time X of the data of the extracted data sub-block n, the data of the extracted data sub-block n, the start time of the extracted data sub-block n+1 X's speech segment data;

(412) splicing the speech segment data of the data of the data sub-block n after extraction, the start time X of the data sub-block n+1 after extraction, then carry out Fourier transform;

(413) Simultaneously, carry out inversion and zero padding processing to the reference speech segment signal of data sub-block n, then carry out Fourier transform;

(414) to the result of step (412) and (413) output, i.e. two vectors, carry out vector dot product operation, then carry out inverse Fourier transform, data clipping to output result successively, obtain reference speech segment relative to data child Correlation calculation result for block n.

The beneficial effects of the present invention are:

The invention utilizes digital signal processing and information processing technology to develop and design a robust detection and identification method that is convenient for software or hardware design and realization. Using the periodic repetition characteristics of illegal FM radio stations, based on the correlation accumulation algorithm, detect the repetition times and time points of illegal FM radio stations, so as to complete the detection and identification of illegal FM radio stations. The detection and identification accuracy is high, and the processing is fast. Has outstanding substantive features and notable progress.

Description of drawings

Fig. 1 is the overall implementation flowchart of the present invention-embodiment.

Fig. 2 is an implementation flowchart of the data preprocessing unit of the embodiment of the present invention.

Fig. 3 is an implementation flow chart of the correlation accumulation calculation unit of the embodiment of the present invention.

Fig. 4 is an implementation flow chart of the correlation calculation fast algorithm of the embodiment of the present invention.

Fig. 5 is an implementation flowchart of the correlation peak detection and recording unit of the embodiment of the present invention.

Fig. 6 is an implementation flowchart of the judgment and decision processing unit of the embodiment of the present invention.

Detailed ways

The present invention will be further described below in conjunction with drawings and embodiments. Embodiments of the present invention include, but are not limited to, the following examples.

The method for detecting and identifying illegal FM radio stations should be easy to realize by using a general commercial hardware platform or a general computer software platform. General-purpose hardware platforms, such as large-scale digital integrated circuits such as DSP (Digital Signal Processor) or FPGA (Field Programmable Gate Array, Field Programmable Gate Array). Common software platforms, such as PowerPC embedded system platform, x86-based Windows operating system platform. Based on a general hardware platform, it has the advantages of strong real-time performance, fast calculation speed, and timely results. Based on a general software platform, the calculation speed is relatively low, but it has the characteristics of greater flexibility and applicability.

Example

As shown in Figure 1, a detection and identification method of an illegal FM radio station comprises the following steps:

S101, sample data collection, carry out continuous voice recording for 3-5 days to the target broadcasting station (it can be a normal broadcasting station or an illegal broadcasting station), the sampling rate is 22050Hz, the bit width of the data sampling point is 16bit, and stored as sample data document.

In the above process, the received target radio signal is an FM broadcast signal, and then the FM broadcast signal is digitally down-converted, decimated and filtered, and the data rate is reduced to 200KHz, and then the FM demodulation process is performed to obtain the original broadcast voice signal , and then perform low-pass filtering and variable rate processing, the data rate is reduced to 22050Hz, and the data sample bit width is 16bit.

S102, sample data preprocessing, the sample data file is divided into multiple short-term data files in units of hours, each file is 158.76MB in size, and is extracted at an extraction rate of 5, that is, every interval of 5 data samples is taken One data sample point, so that the amount of data is reduced to one-fifth of the original, and the size of the data is 31.752MB. The extracted data is stored, and each extracted data is a data sub-block, and these data sub-blocks Name the index number.

For example, the data sub-block with the index number 1 represents the extracted speech data of the first hour period of the entire sample data, and so on.

S103, determine the reference data segment, calculate the data length of the 5-minute speech segment according to the sampling rate and the bit width of the data sample point, and intercept the data sub-block corresponding to the first hour from the starting position, thereby obtaining the first The reference speech segment of the first 5 minutes of the hour period is named as reference signal ref1.

Since the data sub-block is the extracted voice data, the reference voice segment is also the extracted voice data.

S104. Calculation of correlation waveforms. Correlation calculations are performed between the reference signal ref1 and all short-term data files point by point. The correlation calculations are completed using a fast algorithm based on FFT (Fast Fourier Transformation).

Finally, a correlative waveform signal of the reference signal ref1 is obtained, which is named as corr1, and its length is the number of extracted voice data samples in the whole period.

S105. Correlation peak detection, for the correlation waveform signal corr1 of the reference signal ref1, find and obtain its maximum value corr1_max, based on the set correlation peak threshold value 0.25, find the local maximum in the correlation waveform signal whose amplitude value exceeds corr1_max*0.25 peak value and record its position. The number of local maximum peak points is the number of repetitions of the reference signal ref1. The position of the peak point of the local maximum is the position of the repeated speech segment in the entire sample data that is basically identical to that of the reference signal ref1, and the number and position information of these peaks are stored and recorded as the correlation detection record mem1.

S106. For the short-term data segment of the second hour, intercept the voice data segment of the first 5 minutes, and obtain the relevant waveform signals corr2 and mem2 according to the method of steps S103, S104, and S105. For the third hour, the fourth hour, until the last hour of the full-time voice data, the operations of steps S103, S104, and S105 are performed in sequence. Finally, the record information result of the number of repetitions and the position of the first 5 minutes of each hour is obtained.

S107: Accumulate and sum the number of repetitions in the repeated recording information results of all the reference speech segments to obtain a value of the number of repetitions REPEAT. Compare REPEAT with the preset repetition times threshold value Gate_rp. The value of Gate_rp is 8. If it is greater than the threshold value, it is determined that the repetition rate of the full-time voice data is very high, and the target broadcasting station is an extremely suspicious illegal broadcasting station.

S108: Repeat steps S101-S107 for all monitored target broadcast stations to obtain judgment results of all monitored target broadcast stations, thereby automatically screening out highly suspected illegal broadcast stations. The location information in the mem is recorded through the relevant detection of each radio station, and the repeated location can be targeted for selective playback and playback of voice clips for further verification.

Figure 2 is the flow chart of the implementation of the sample data preprocessing unit, combined with Figure 2, the specific implementation details are described as follows:

S201: For a certain target broadcasting station, use the receiver to receive and collect the intermediate frequency signal, then perform DDC down-conversion and FM demodulation processing, and finally obtain a long-term voice signal sample data with a sampling rate of 22050Hz and a 16-bit bit width; usually The length of time is 3-5 days.

S202: Calculate the length of the sample data, intercept the sample data of an integer hour, and discard the last non-integer hour data. The entire sample data is divided into several data sub-blocks in units of hours, and the speech length of each data sub-block is 1 hour. For example, for 3 days of sample data, 72 data sub-blocks are finally obtained.

S203: Perform a 5-fold data extraction operation for each data sub-block. This is very necessary to reduce computing time without affecting the final decision result. Therefore, the data volume of each data sub-block will be one-fifth of the original. The amount of data sub-blocks for one hour is also reduced from 158.76MB to 31.752MB.

S204: Store the extracted data sub-blocks as files, and number them in chronological order; these extracted data sub-blocks are used as processing objects in the subsequent processing flow.

Fig. 3 is the realization flow chart of relevant cumulative calculation unit, combined with the specific realization details of Fig. 3, the description is as follows:

S301: Set N as the number of data sub-blocks. Set the initial value of n to 1, obtain the data of data sub-block n, and intercept the voice data of the initial 5 minutes as the reference voice segment ref_n.

S302: Obtain all the data of data sub-block n, obtain the voice data of the first 5 minutes of data sub-block n+1, and splice the two to obtain sample data that needs to be compared with the reference voice segment. The reason for splicing is to ensure the integrity of the result of the correlation operation between the reference speech segment and the speech data of the data sub-block n. The number of sample points of the reference speech segment is L1, the number of data sample points of the data sub-block n is L2, and L2 is much larger than L1. When correlation calculation is performed on the end data of , the initial part of data of data sub-block n+1 is needed. The length of this part of data is just equal to the length of the reference speech segment, so the initial 5-minute data of sub-block n+1 is needed.

S303: Carry out correlation calculation based on the fast correlation algorithm for the reference speech segment ref_n and the data sub-block sample data b_n, and obtain a correlation calculation result corr_bn_rn. For the fast correlation calculation algorithm, please refer to Figure 4 and its description.

S304: Accumulate n by 1, and judge whether n is equal to the total number N of data sub-blocks, if less than N, return to step S302. If it is equal to N, then end, and splicing all the correlation calculation results into a complete data, this data is the correlation calculation result of the reference speech segment ref_n relative to the entire sample data - the correlation waveform signal corr_n.

Figure 4 is an implementation block diagram of the fast algorithm for related calculations, combined with the specific implementation details of Figure 4, the description is as follows:

First obtain the processing object of correlation calculation, namely the reference speech segment signal (length is L1) of module 401, the data sub-block n data (length is L2) of module 402, the data of the data sub-block n+1 of module 403. Module 405 completes the splicing of data sub-blocks. The splicing method has been described in detail in the description of the implementation process in FIG. 3 and will not be described again. The length of the spliced sub-block data is L3=L1+L2; for the reference speech segment signal, the module 404 completes the inversion and zero padding of the signal. After inversion and zero padding processing, the last data sample point of the reference speech segment signal ref_n becomes the first data sample point, and then L2 zeros are added, and finally the first to L1-1th of the reference speech segment signal sample points. Fast Fourier Transform (FFT: Fast Fourier Translation) is performed on the concatenated data sub-block data and the reversed reference speech segment, that is, the processing operations completed by module 407 and module 408 . The output results of module 407 and module 408, that is, the result of performing fast Fourier transform, correspond to two vectors, and perform vector dot product operation. Both vectors have the same dimension, both L3. The so-called dot product means that the corresponding elements of the vector are multiplied, and the result is still a vector with dimension L3. The dot product operation of these two vectors is completed by module 409 . Module 410 completes the inverse Fourier transform, which is also realized by using the fast Fourier transform algorithm. Module 411 completes the clipping of the data, the length of the output vector of the inverse Fourier transform is L3, and the first L2 data is intercepted to obtain the correlation calculation result of the reference speech segment relative to the data sub-block n.

Fig. 5 is the realization flowchart of correlation peak detection and recording unit, in conjunction with the concrete realization detail description of Fig. 5 as follows:

S501. Obtain a correlation calculation result corr_n of the reference speech segment n relative to the entire sample data.

S502. Find the global maximum peak point from the corr_n corr_n correlative waveform signal, and set the peak value as corr_n_max.

S503. Determine the local correlation peak threshold value as gate_corr=corr_n_max*0.25.

S504. In the entire correlation waveform signal corr_n, search for all local maximum peak points exceeding gate_corr, determine the number and positions of these local peaks, and store and record them as data mem_n.

Fig. 6 is the realization flowchart of judgment decision-making processing unit, in conjunction with the concrete realization detail description of Fig. 6 as follows:

S601. Set N as the number of reference speech segments, that is, subtract 1 from the total number of data sub-blocks.

S602. Set the initial value of the REPEAT value to 0.

S603. Set n=1, and obtain the correlation peak detection result mem_n of the reference speech segment n relative to the correlation calculation result corr_n of the entire sample data.

S604. Obtain the number of correlation peaks repeat_n from mem_n, and record it as the number of repetitions of the reference speech segment n in the entire sample data; and set REPEAT=REPEAT+repeat_n.

S605, add 1 to n, if it is less than N, return to the previous step to obtain the number of repetitions of the new reference speech segment in the entire sample data, and accumulate REPEAT until n is equal to N.

S606. Finally, obtain the sum of the number of repetition records of all reference speech segments relative to the entire sample data, and record it as REPEAT.

S607. Compare REPEAT with a preset detection threshold value, and if it is greater than the threshold value, determine that the target broadcasting station is an illegal station; otherwise, it is a regular station.

The division of modules or devices provided in the embodiments of the present invention is only functional division, and may be physically combined or divided. The names of modules or devices may change with technological evolution or different application scenarios, but should still be considered within the protection scope of the present invention. Those of ordinary skill in the art can understand that all or part of the steps in the method of the above-mentioned embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When executed, several instructions are included to execute the methods described in various embodiments of the present invention. The storage medium mentioned here, such as: ROM/RAM, magnetic disk, optical disk, etc.

According to the above-mentioned embodiments, the present invention can be well realized. It is worth noting that, based on the premise of the above-mentioned design principle, in order to solve the same technical problem, even if some insubstantial changes or modifications are made on the basis of the structure disclosed in the present invention, the essence of the adopted technical solution is still Like the present invention, it should also be within the protection scope of the present invention.

Claims (8)

1. A detection and identification method of an illegal FM radio station, characterized in that, comprising the steps: (1) Sample data collection, continuous voice recording of the target broadcasting station, and store as a sample data file; (2) Sample data preprocessing, divide the sample data file into multiple data sub-blocks, and number and store each data sub-block; (3) Determine the reference data segment, and sequentially acquire the voice segment data of the start time X of each data sub-block as the reference voice segment; (4) Calculation of correlation waveforms, correlating each reference speech segment with the data of all data sub-blocks point by point to obtain the correlation waveform signal of each reference speech segment; (5) Correlation peak detection, for each reference waveform signal of the reference speech segment, find and obtain the maximum value, and find the local maximum peak value whose amplitude value exceeds the maximum value * preset correlation peak threshold value in the relevant waveform signal point, and record its position; (6) Accumulate and sum the number of times in the repeated recording information results of all reference voice segments to obtain the number of repetitions, and then compare the number of repetitions with the preset threshold value of repetitions to determine whether the target broadcasting station is is an illegal broadcasting station; (7) Repeat steps (1)-(6) for other target radio stations. 2. A method for detecting and identifying illegal FM radio stations according to claim 1, characterized in that, in the step (1), the continuous voice recording time is 3-5 days, and the sample data sampling rate is 22050Hz, the data sample bit width is 16bit. 3. A method for detecting and identifying illegal FM radio stations according to claim 2, characterized in that in the step (1), the sample data collection process of the target broadcast station is as follows: first, receiving the target broadcast station FM broadcast signal, and then carry out digital down-conversion and decimation filtering on the FM broadcast signal, and reduce the data rate to 200KHz, and then perform FM demodulation processing to obtain the original broadcast voice signal, and then perform low-pass filtering and variable rate processing, and the data The rate is reduced to 22050Hz, and the data sample bit width is 16bit. 4. A method for detecting and identifying illegal FM radio stations according to claim 3, characterized in that, in the step (2), the sample data file is divided into multiple data sub-blocks in units of hours . 5. A detection and identification method for an illegal FM radio station according to claim 4, characterized in that, each data sub-block after the step (2) is further divided into data sample points, and the The extracted data sample points are indexed and named according to the number of the data sub-block, and stored as a new data sub-block. 6. A method for detecting and identifying illegal FM radio stations according to claim 5, characterized in that, in the step (3), first, the time X is calculated according to the sampling rate, data sample bit width, and extraction rate The data length of the voice segment, and then according to the data length, the voice segment data of the start time X of each data sub-block is obtained as a reference voice segment. 7. A detection and identification method for an illegal FM broadcasting station according to claim 6, characterized in that, in the step (4), the correlation calculation adopts fast Fourier transform calculation, and the specific calculation process is as follows: (401) First select the reference speech segment ref that needs to be correlated. There are N data sub-blocks in total, so there are N reference speech segments, n≤N, and the initial value of n is 1; (402) Obtain all the data of the data sub-block n, obtain the voice segment data of the start time X of the data sub-block n+1, and splicing the two to obtain the sample data b_n that needs to be compared with the reference voice segment; (403) Carry out correlation calculation on the reference speech segment ref and b_n, and obtain the correlation calculation result corr_bn; (404) Accumulate n by 1, and judge whether n is equal to the total number of data sub-blocks N, if it is less than N, then return to step (402); if it is equal to N, then end, and put all related calculation results corr_bn, n is 1 An integer between N and N is spliced into a complete data, which is the correlation calculation result of the reference speech segment ref relative to the entire sample data - the correlation waveform signal corr. 8. A method for detecting and identifying illegal FM radio stations according to claim 7, characterized in that in the step (403), the correlation calculation is based on fast Fourier transform calculation, and the processing of fast Fourier transform The objects are all extracted data, and the specific process of related calculations is as follows: (411) Obtain the processing object, that is, the voice segment data of the starting time X of the data of the extracted data sub-block n, the data of the extracted data sub-block n, and the starting time of the extracted data sub-block n+1 X's speech segment data; (412) Splicing the data of the extracted data sub-block n and the speech segment data of the starting time X of the extracted data sub-block n+1, and then performing Fourier transform; (413) At the same time, perform inversion and zero padding processing on the reference speech segment signal of the data sub-block n, and then perform Fourier transform; (414) Perform vector dot product operation on the output results of steps (412) and (413), that is, two vectors, and then perform inverse Fourier transform and data clipping on the output results in order to obtain the reference speech segment relative to the data sub- Correlation calculation result for block n.