KR100456408B1 - Search of audio date and sample - Google Patents

Abstract

The present invention receives audio displayed on a radio or television from a wireless communication device such as a mobile phone, and generates an audio gene used for audio recognition as a gate, and recognizes the audio using the audio gene. It relates to a data retrieval method.

Description

Audio Gene Generation Method and Audio Data Search Method {SEARCH OF AUDIO DATE AND SAMPLE}

The present invention relates to a method for generating an audio gene and a method for retrieving audio data, and more particularly, to receive an audio displayed on a radio or a television from a wireless communication device such as a mobile phone, and to generate an audio gene used for audio recognition by using a voiceprint. An audio gene generating method and an audio data retrieval method for recognizing audio using the audio gene.

There is an increasing need for automatic recognition of music or other audio signals originating from various sources. That is, music copyright holders want to obtain as a basis for how much their music is broadcasted on radio or television and how much can be claimed. Also, with the development of digital data such as MP3, when you download music files through the Internet, you want to download music files from the radio or television.

A number of methods for automatically recognizing audio signals using a computer system have been introduced in the past, but are not currently used. One of the automatic recognition methods was a "search method in an audio database" (Korean Patent Publication No. 2003-59085), which was patented in Korea on July 7, 2003.

The search method in this audio database (Korean Patent Publication No. 2003-59085) includes determining a set of fingerprints at a specific location of a sample, finding a matching fingerprint in a database index, and location of the sample. Creating a correspondence between the locations of the files having a fingerprint equal to and identifying a media file to which a substantial number of correspondences are substantially linearly related. One way to identify files with a large number of correspondences has been to perform the equivalent of scanning the diagonals of the scatter plot generated from pairs of correspondences.

Fig. 8A shows a linear correspondence between positions having the same fingerprint between the sample and the audio data, so that the audio data contains a sample, and Fig. 8B shows no linear correspondence, so no audio data containing the sample was found. It is shown.

The search method in this audio database selects audio data in which the correspondence between a sample having the same fingerprint and the position of the audio data is linearly related to the sample, and has a higher level of noise and distortion than the previous methods. It has been effective in screening the received audio signal in relatively real time.

However, in the conventional audio database, a search method must first select all fingerprints of the same audio data as the fingerprints of the sample, and then check the linearity of the positional relationship between the audio data and the sample having the same fingerprint for each audio data. Therefore, the search time was not faster than the search method originally expected. In other words, the same fingerprints need to be selected to confirm the positional relationship, which causes a waste of time.

In addition, in the conventional audio database search method, since a landmark, which is a position that can be calculated to be reproduced, must be stored together with each fingerprint in order to compare the audio data and the sample, the database of the audio data There was a problem in that the capacity of and the size of the list of samples became large. This increase in capacity and size has a problem in that the overall processing speed of the processor lowers the search speed sequentially.

In addition, the conventional method of searching in the audio database suggests a method of recognizing all audio data, but specifically presents a method of searching for a modified audio signal in which the conversion or modulation of the audio signal exists, such as a mobile phone or a PCS phone. I did not want to leave.

An object of the present invention is to solve the above problems, and to search for audio data using a fingerprint, the audio gene generation method and audio that can improve the speed of searching the audio data by not storing the location information of the fingerprint separately It is to provide a data retrieval method.

In addition, another object of the present invention is to provide an audio gene generation method and an audio data retrieval method capable of recognizing audio data of a modulated audio signal such as a mobile phone or a PCS phone.

Another object of the present invention is to provide an audio gene generation method and an audio data retrieval method capable of minimizing the capacity of data to be compared or stored, thereby saving storage space and improving processing speed of a processor.

In addition, another object of the present invention is to extract the frequency from the overlapping time intervals to compare the sample, even if an error occurs in one time interval to compare and search because overlapping can improve the search rate, that is, the accuracy of the search The present invention provides a method for generating an audio gene and a method for retrieving audio data.

1 is a conceptual diagram of a system for searching audio data according to an embodiment of the present invention;

2 is an overall flowchart of a method of retrieving audio data according to an embodiment of the present invention;

3 is a flowchart of an audio sample generation step of the mobile phone of FIG. 2;

Fig. 4a is a graph of the signal magnitude during the sample time of the sample.

FIG. 4B is a graph of signal magnitudes of frequencies included in a particular time interval of FIG. 4A. FIG.

Figure 4c is a graph of the signal size of Figure 4b amplified when more than a certain size, attenuated when less than.

5 is a flow chart of generating a sample audio gene from the sample of FIG.

6 is a flowchart of a step of comparing the sample audio gene and audio data audio gene of FIG.

Fig. 7 is a diagram showing frequencies of piano scales which is a reference frequency used to remove noise.

Fig. 8A is an example of the conventional search method, which shows a linear correspondence between positions having the same fingerprint between the sample and the audio data so that the audio data includes the sample.

FIG. 8B is an example of the conventional search method, in which no audio data including a sample was found because no linear correspondence was shown. FIG.

Explanation of symbols on the main parts of the drawings

10: audio data retrieval system 12: information and communication equipment

14: ARS System 16: Audio Generating Server of Sample

18: Audio data storage DB or audio information storage DB

20: search server 22: radio

24: sample storage server 10: audio sample storage step

S20: Audio Gene Generation and Storage Step of Sample

S30: Audio data DB generation step S40: Audio gene search step

S50: Search result output step

In order to achieve the above object, the present invention provides an audio gene generating method for generating an audio gene from audio data, comprising: a time division step of dividing an audio signal at regular time intervals; A frequency conversion step of calculating a magnitude of a signal of frequencies included in each time interval or in a plurality of time intervals; Calculating a difference in the magnitude of the signal between adjacent frequency sections by dividing the frequency domain into predetermined sections; A slope calculation step of obtaining a difference of said calculated value between adjacent time intervals; A quantization step of quantizing to 1 when the slope is greater than or equal to 0 when less than 0; And generating an audio gene by storing the quantized values.

In another aspect, the present invention is an audio data retrieval method for retrieving the same audio data as the audio sample by comparing the audio sample storing the audio signal for a predetermined time and a plurality of audio data stored the entire audio signal, A sample audio gene generation step of generating an audio gene of an audio sample using the audio gene generation method of claim 1; An audio data audio gene generating step of generating the audio data of the audio data using the audio gene generating method of claim 1; And a search step of searching for audio data including the same audio gene as the sample audio gene.

The searching may include selecting a predetermined number of sample audio genes from the sample audio genes; A pseudo value generating step of generating a bit value that is the same as or different from the one sample audio gene; A section searching step of searching for a section having the same value as the similar value among the audio data audio genes, and searching for the audio data audio gene having the same interval as the selected section of the sample audio gene; It may have a selection step of calculating whether the audio data including the similar value at the same interval and the sample audio gene is the same as a whole, and selecting only the same audio data as the audio sample when the difference is less than a predetermined standard. .

In addition, the audio sample may be an audio signal transmitted from a wireless communication device from which noise is removed since amplification of a signal size greater than a predetermined signal size and attenuation of a value less than a predetermined signal size are reduced from the audio signal input from the wireless communication device.

In addition, the audio sample may be an audio signal that is directly connected and transmitted from an audio apparatus such as a radio or a television.

Further, in the searching step, the audio genes of the audio data of the database in which the audio data are stored are distributed and temporarily stored in a distributed system, and the search time is sequentially or simultaneously compared when the sample audio gene and the audio genes of the audio data are compared. Can shorten.

In addition, the searching step may further include a broadcast counting step of storing the number of selections of the audio data after the selecting step to calculate the number of times broadcast from the audio device.

In the present specification, "sample" or "audio sample" is output from media or audio devices such as radio or television, recorded recordings, commercials, MP3 players, CD players, indoor or outdoor performances, event music, logo songs, music rhythms, etc. A piece of audio data of any size obtained from all kinds of audio, including sound, voice, music, or a combination thereof.

In addition, in this specification, "audio data" means all kinds of audio data including sound, voice, music, or a combination thereof.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Techniques or facts well known to those skilled in the art in the description of the embodiments will be omitted and omitted.

Example 1 Audio Data Retrieval System

Referring to FIG. 1, the audio data retrieval system 10 according to the first embodiment of the present invention includes an information communication device 12, an ARS system 14, an audio gene generation server 16 of a sample, and an audio data storage DB. Or an audio information storage DB 18 and a search server 20 for samples and audio data. The audio data retrieval system 10 also has a radio 22 and a sample storage server 24.

The information communication device 12 includes all terminals capable of connecting to the ARS server 14 using a communication network such as a mobile phone, a telephone, a radio, a terminal equipped with a wired / wireless communication modem. The information communication device 12 transmits audio data such as music or sound source output from a radio or TV, a performance site, or an event site to the ARS server 14 through a communication network. At this time, the type and form of audio data includes not only existing ones but also all things newly developed in the future, and the media on which audio data is output are not only radio or TV, but also any media media, performances, event events, performances online or offline. Include things.

The ARS server 14 allows the information communication device 12 to select a type of service when making a connection request, and if audio data is transmitted from the information communication device 12 according to the service type, the ARS server 14 stores the audio data primarily. Function

The audio gene generation server 16 is connected to the ARS server 14 by a communication network, and is suitable for comparing audio data stored in the ARS server 14 with audio genes already stored in the audio information storage DB 18. That is, it converts to the same form and transmits to the audio gene search server 18. In this case, the audio gene may be an audio DNA using audio fingerprinting from the audio data. In this case, the audio gene may be an audio gene using a voiceprint of the modulated or converted audio data when data modulation or conversion is performed by sampling in an environment in which noise or sound distortion occurs according to the transmission method of the information communication device 12. . For example, when the information communication device 12 is a mobile phone, the noise and the call tone are distinguished and filtered to be considered noise. Thus, only the remaining filtered signals are used as the voiceprint of the audio data to generate an audio gene. The method of generating the audio gene is described in detail in the second embodiment.

The audio information storage DB 18 is a DB in which various audio genes are stored in advance from a music or a sound source. At this time, the stored audio genes are audio genes generated by using a voiceprint of audio data modulated or converted according to the type of the information communication device 12. In this case, the audio gene is also sampled in the environment in which noise or sound distortion is generated from the information and communication device 12, and in order to compare the audio gene generated by the data modulation or the converted audio data into the gate, the modulated or converted audio in advance Create and store audio genes using the gates of the data.

Audio genes stored in the audio information storage DB 18 are classified based on the composition, time signature, and numerical value of the corresponding audio data. The method of generating the audio gene is preferably the method of generating the audio gene generating server 16. However, if the same audio gene can be searched by comparing the two audio genes, another generation method may be used.

In this case, the audio information storage DB 18 stores not only the audio gene but also music information corresponding to the music gene, for example, information related to audio data such as a song name, a singer, lyrics, and an advertiser.

The method of generating and storing the audio gene of the audio data stored in the audio information storage DB 18 will be described in detail in the second embodiment. The audio information storage DB 18 generates the original sound audio gene DB (indicated as DNA DB in FIG. 1), which generates and stores the audio gene as the voice text, and the pre-modulated audio signal when the original sound is modulated through a mobile phone. Has a modulated audio gene DB (denoted as fDNA DB in FIG. 1) that generates and stores an audio gene. On the other hand, the audio information storage DB is managed by a normal DBMS 21.

The audio gene search server 20 compares a plurality of audio genes previously stored in the audio information storage DB 18 with audio genes transmitted from the audio gene generation server 16, and matches the corresponding audio genes and other related music information. Function to find. At this time, the audio gene search server 20 narrows down the search by using the composition, time signature, and numerical value before directly comparing the audio genes in order to efficiently search for the audio genes from the audio information storage DB 18. ?Tolerance? The technique is applied to the search method to quickly search for patterns of audio genes. That is, if the pattern of the audio gene to be searched and the pattern of the audio gene of the audio information storage DB 20 are searched and show an error within an allowable error, for example, 36%, the same audio gene is considered.

The audio gene search server 20 includes a plurality of search servers in order to improve the search speed. The audio gene search server 20 divides and stores the audio genes from the audio information storage DB 18 in each search server, and each search server simultaneously or It is possible to sequentially search for the same audio gene as the audio gene transmitted from the ARS server 14.

For example, if there are 100,000 audio information stored in the audio information storage DB 18 and the audio gene search server 20 is composed of 10, each search server 20 stores only 10,000 audio information, respectively. When the audio gene is transmitted from the ARS server 14, the audio gene may be searched simultaneously or sequentially. This can improve the search speed of audio genes by 10 times.

A comparative search method using audio samples generated from the audio samples and audio data of the audio gene search server 20 will be described in more detail in the second embodiment.

Example 2 Audio Data Search Method

The audio data retrieval method according to the second embodiment of the present invention includes the audio sample storage step (S10), the audio gene generation and storage step (S20) of the sample, the audio data DB generation step (S30), and the audio gene search step (S40) and A search result output step S50 is provided.

1 to 4, the audio sample storage step (S10) is a sample time of the audio signal input through the information communication device, for example, the mobile phone 12 by the speech coder (speech coder) of the mobile phone 12 (3 seconds) is stored as a sample, and starts from the step of measuring the size (db) of the audio signal over time as shown in Figure 4a (S12). In this case, the sample measures and stores the size of the audio signal according to a frequency in the range of 300 Hz to 3 kHz every 11 msec for 3 seconds (S14).

As shown in FIG. 4B, an audio signal having a specific size, for example, 500db or more, is amplified by 10 times, and attenuated by 0.1 times by an audio signal having a size of less than 500 db (S16). ). As shown in FIG. 4C, the audio signal is converted into a form in which only a signal of a specific frequency exists every 11 msec (S18). The converted audio signal is stored in the form of an audio file such as A.WAV and transmitted to the ARS system 14 through the mobile phone 12. Therefore, the audio sample includes an audio signal amplified or attenuated and modulated among the audio signals in the frequency range of 300 Hz to 3 kHz for 256 * 11 msec in 11 msec units.

1, 2, and 5, the step S20 of generating a sample audio gene from an audio sample is performed by the sample audio gene generation server 16 connected to the ARS system 14. It starts by storing the audio sample generated by S10).

Audio samples are divided into 33 frequency sections (FI1 to FI33) by selecting only frequencies of 750Hz to 2750Hz. The size of the audio signal according to each frequency section is stored separately. As a result, as shown in Equation 1, 256 * 33 audio signals corresponding to the product of 256 time intervals (unit time: 11 msec) and 33 frequency intervals (unit frequency: 66 Hz) are stored (S21).

Magnitude of the signal [i, j] = [Ai, j]

Here, i (natural number of 1 ≦ i ≦ 256) means 256 time intervals, and j (natural number of 1 ≦ j ≦ 33) means 33 frequency intervals.

In order to generate an audio gene from the audio signal, the difference in signal magnitudes between adjacent frequencies of a specific time period, for example, 11 msec, for example, FI1 and FI2, is obtained as shown in Equation (2). The difference in the magnitudes of the signals of FI2 and FI3, FI3 and FI4, ... FI32 and FI33 is calculated in the same way. In the next time period, 11 * 2msec, the difference in signal magnitudes between adjacent frequency sections is obtained. In the same way, obtain up to 256 * 11msec. Therefore, the value of the difference of the magnitude of 256 * 32 signals is calculated (S22).

Difference in signal magnitude (i = 1) = [A1, k]-[A1, k + 1]

Where k is a natural number from 1 to 32

The difference value calculated as in Equation 3 is calculated by recalculating the difference (hereinafter, referred to as slope) of the difference value between adjacent time zones, for example, 11 msec and 11 * 2 msec time zones, and when the value is 0 or more, it is "1" and less than 0. If so, the value "0" is stored (S23 to S26). This process is performed for the entire time interval (S27). If all the slopes are obtained and stored, 32 * 256 quantized binary values are stored (S28). This quantized binary value is called the audio DNA of sample.

Slope = ([A1, k]-[A1, k + 1])-([A2, k]-[A2, k + 1])

Where k is a natural number from 1 to 32

Referring again to FIGS. 1 and 2, the method S30 of generating the audio gene of the audio data stored in the audio data DB 18 is generated in the same manner as the method of generating the audio gene of the sample. However, although the sample has a sample time of about 3 seconds, the audio data generates an audio gene for all audio data. The audio genes of the entire audio data are stored in the audio data DB 18 and managed by the DBMS 21.

1, 2, and 6, a search step (S40) of the sample audio gene and the audio gene of the audio data is the sample audio gene and audio generated by the sample audio gene generation server 20 through the mobile phone 12. Comparing the audio genes of the audio data stored in the data DB to find an audio gene containing a sample audio gene. This step is performed simultaneously or sequentially in ten distributed search servers 20 which are distributed from the audio data DB 18 and temporarily store the audio gene of each audio data.

The searching step S40 of the audio gene may include selecting step S42 of selecting a predetermined number of sample audio genes from sample audio genes, and generating a similar value generating a bit value that is the same as or different from the sample audio gene by one bit ( A section search step (S46) of searching for a section having the same value as the similar value among the audio data audio genes, but searching for the audio data audio gene having the same interval as the selected section of the sample audio gene; A final selection step (S46) is performed to calculate whether the audio data and the sample audio gene are similar to the audio data including similar values at equal intervals, and to select the same audio data as the audio sample only when the difference is less than a predetermined standard. Have

In the selection and similarity generating step (S42, S44), first of 11 msec of the sample audio genes, each of the audio genes of six time zones of 11 * 50msec, 11 * 100msec, 11 * 150msec, 11 * 200msec, and 11 * 250msec By generating bit values that are the same or only one bit out of the bits, audio data audio genes having the same bit values are retrieved from each distributed search server 11.

In the section search step (S46), a section having an audio data audio gene maintaining a distance of 11 * 50 msec among audio data audio genes having the same bit value of the same or only one bit among 32 bits in six time zones is found. Serve If this section does not exist, audio genes of other audio data are sequentially searched.

In the final selection step (S48), when a section is searched, a bit error rate (BER) is calculated by comparing the entire sample audio gene and the audio gene of the budget section among the audio genes of the audio data. If the BER of the entirety of 256 sample audio genes and the predicted interval audio data is a constant value, for example, 32% or less, it is determined that the audio gene of the audio data includes the sample audio gene and the distributed server. Terminate search. Of course, if the BER exceeds 32%, the same search is performed sequentially for the other audio data.

Referring again to FIGS. 1 and 2, the information of the audio data including the sample audio gene is output using the audio information stored in the audio data DB 18. All the procedures are completed by searching the same audio data as the sample and outputting the result.

Example 3 Radio Sample Collection

If the sample storage server 24 can directly capture and store the audio sample from the radio 22, since the sound source of music can be stored as it is, the sample audio gene is generated from the sound source and transmitted to the search server 20. The search server 20 performs a comparative search using the audio gene while temporarily storing the audio gene of the audio data generated from the original sound of the music temporarily stored from the audio data DB 18. In the same manner as described above, the audio data including the sample audio gene is determined by comparing between the sample audio gene and the audio data audio gene. The search results are output, so the series of processes ends.

Example 4: Noise Reduction

In the present invention, after an audio sample of a modulated audio signal is transmitted from the mobile phone 12 described in Embodiments 1 and 2 and stored in the ARS system 14, the audio sample is generated before generating a sample audio gene from the audio sample. It provides a method for removing the noise included in the.

As described in Embodiments 1 and 2, the audio samples transmitted from the mobile phone 12 and stored in the ARS system 14 are already amplified or attenuated according to a predetermined size of the audio signal in the mobile phone 12, and thus have a frequency of 300 Hz. Audio signal modulated in the 3kHz range. However, this audio sample contains unnecessary noise as well as an audio signal extracted from audio data such as music or sound source.

One of the methods of removing noise included in the audio sample includes (1) a non-existent frequency removing step of removing a signal having a certain size or less in the frequency domain of the audio sample, and (2) among the frequencies of the remaining audio sample. It has a meaningful frequency selection step of selecting only meaningful frequencies, that is, frequencies of an audio signal extracted from audio data such as music or sound source.

In the non-existent frequency removing step, signals in a frequency domain having a signal size less than or equal to a predetermined size among audio samples are regarded as noises generated by interference of the audio signal and removed. Since the mobile phone 12 amplifies and attenuates the signal based on the magnitude of the signal, noise is removed to some extent, but once again, the noise is removed by the magnitude of the signal.

Since the mobile phone 12 is a main purpose of the call, even if a certain range of noise is included, since there is no problem in the call, there is no choice but to include a lot of noise. In the present invention, the noise is once again removed based on the size of the signal for audio data retrieval. However, if the mobile phone 12 is removing or removing noise based on the signal size to the extent necessary for the practice of the present invention, this non-existent frequency removing step may be omitted.

Significant frequency selection step is focused on the fact that the audio data processed in the present invention is mainly an audio signal extracted from a music or a sound source, such that the frequencies that do not exist in the music or sound source, etc. are considered as noise even if the signal size is above a certain reference value. It is a step to remove. One method of selecting frequencies existing in music, sound sources, and the like is to select only signals corresponding to a frequency range within a predetermined range from the frequencies of the scales based on the frequencies of 96 piano scales.

As can be seen from Fig. 7, if there are 96 piano scales, each scale has a specific frequency. For example, the lowest scale, C1, is 32.70 Hz and the highest scale, B8, has 7,900.88 Hz. Therefore, the audio data extracted from the music or the sound source, etc. may be audio signals composed of frequencies equal to or within a predetermined range of the piano scale.

The allowable frequency range based on the frequency of the piano scale can be easily set by those skilled in the art. For example, if the allowable frequency range is set to 10% of the frequency of the piano scale, the frequency range of the C3 scale can be set to 130.80 ± 13.08Hz.

Since only the audio signal in the allowable frequency range of the frequency of the piano scale is selected from the audio sample transmitted from the mobile phone 12 and stored in the ARS system 14, most noises such as human sounds, car sounds, and wind noises can be removed.

Example 5 Method of Using Selected Audio Data Information

The present invention relates to audio data information selected by the systems and methods described in the first to third embodiments, for example, music information such as song names, lyrics, and singers, cell phone ringtones, call waiting sounds, and voice messages using the retrieved music. Additional services such as background music can be provided via mobile phone or wired / wireless internet.

In addition, the present invention may be used to determine the music copyright by providing broadcast information recorded by counting the number of broadcasts of music output from a TV or a radio to a broadcasting station, a record sales company, a record production association, etc. using a communication network.

In addition, the present invention transmits audio data such as music through mobile phones or the like described in Patent Application No. 2004-1246 (Invention Name: Advertising Method using Audio Data Recognition), which is filed by the present inventors, to an event such as a prize. Participation may be used in an advertising method using audio data recognition, which can maximize the effect of an advertisement including audio data.

Hereinafter, the operation and operation of another embodiment according to the present invention described above will be described in detail.

1 to 7, if the user wants to hear the music heard from the radio 22 and to know the information about the music, the mobile phone 12 connects to the ARS system 14 by telephone. The ARS system 14 instructs the radio on which music is output to face the mobile phone for about 5 seconds. At this time, the mobile phone 12 divides the music and various noises inputted into time zones by using a voice coder (VOCODER or speech codec) built in the mobile phone 12 and divides them by frequency and amplifies when the signal size is 500db or more, and attenuates it when it is less than. To the ARS system 14.

The audio gene generation server 14 obtains slopes for the modulated audio signals, quantizes the slope values, and stores 256 sample audio genes in 32 bits. The stored sample audio gene is transmitted to the search server 20 distributed in the form of a wav file, and the audio data DB or the audio data of the audio information DB 18 including the same audio gene by the search servers 20 at the same time. Search for the audio gene.

In retrieval, the element values of six audio genes of the sample audio gene are selected at intervals of 11 * 50 msec so that element values equal to or different from these element values are transmitted to the audio genes of the audio data through the respective search server. Will be searched for. Once the same six element values are retrieved, only those with 11 * 50msec spacing are selected. The interval between the selected audio genes and the entire sample audio gene are compared again and the difference is stored as BER. If the value is less than 32%, the same audio gene is determined, otherwise the same is determined.

If it is determined that the same audio gene as the sample audio gene exists, the search server 20 transmits the result to the DBMS 21, and the DBMS 21 transmits the audio information of the corresponding audio data from the audio data DB 18. You can output the song name, singer, score, lyrics, etc. and send it via SMS on your mobile phone.

Similarly, when the music of the radio 22 is directly stored in connection with the radio 22, the original sound is stored as a WAV file without being tampered with, and the search server 20 searches the same way as the mobile phone 20. The difference is that the audio genes are generated from the original sound, so that the audio genes of the audio data stored in the audio data DB 18 also generate and store the audio genes from the original sound. If the same audio data is retrieved, the audio information as well as the audio information such as music or advertisement output from the radio can be delivered to the copyright holder or advertiser for the original music and used for calculating the copyright fee or for calculating the advertising cost from the advertiser. have.

Although the present invention has been described with reference to the above embodiments, the present invention is not limited thereto.

In the above embodiment, the numerical values such as the 11 msec time interval or 750 to 2750 Hz are not limited to the present invention, and various changes can be made according to the user's requirements such as the type or form of the input audio signal, desired processing speed, and search accuracy. Do.

In addition, in the above embodiment, the audio gene is generated by extracting a frequency from one time section of the audio sample. However, the audio gene may be generated by extracting a frequency by overlapping the frequency from several time sections. As the frequency is extracted by overlapping the time intervals, the presence or absence of a signal according to the frequency can be repeatedly compared and searched.

In addition, in the above embodiment, although the audio signal of all frequencies is extracted from one time period or several overlapping time intervals, an audio gene is generated from all the extracted audio signals. You can also extract and generate audio genes only from them. Since the audio signals to be searched use only the audio scale corresponding to the specific scale of the east and west, that is, the specific frequency, even if the audio signal is generated by extracting the audio signal only from this specific frequency, there is no problem in the comparative search of the audio data. There is an effect that can significantly shorten.

In addition, in the above embodiment, the components have been described by dividing into various servers, but the functions of the servers of the above embodiment may be performed by several modules or programs in one server, or a function performed by one server. You can also run it on multiple servers. For example, the audio gene generation server and the audio gene search server have been described as being divided, but one function may be performed by one server.

Audio gene generation method and audio data retrieval method according to the present invention, it is possible to improve the search speed of the audio data by retrieving the audio data using the fingerprint, but does not store the position information of the fingerprint separately.

In addition, the audio gene generation method and the audio data retrieval method according to the present invention have the effect of recognizing the audio data of a modulated audio signal such as a mobile phone or a PCS phone.

In addition, the audio gene generation method and the audio data retrieval method according to the present invention have the effect of saving the storage space and improving the processing speed of the processor by minimizing the capacity of data to be compared or stored.

In addition, the audio gene generation method and the audio data retrieval method according to the present invention extract the frequencies from the overlapping time intervals of the samples to be compared. There is an effect to improve the accuracy of.

Claims (8)

An audio gene generation method for generating an audio gene from audio data, A time division step of dividing the audio signal at predetermined time intervals; A frequency conversion step of calculating a magnitude of a signal of frequencies included in each time interval or in a plurality of time intervals; Calculating a difference in the magnitude of the signal between adjacent frequency sections by dividing the frequency domain into predetermined sections; A slope calculation step of obtaining a difference of said calculated value between adjacent time intervals; A quantization step of quantizing to 1 when the slope is greater than or equal to 0 when less than 0; And generating an audio gene by storing the quantized values. An audio data search method for searching audio data identical to an audio sample by comparing an audio sample storing an audio signal for a predetermined time with a plurality of audio data storing the entire audio signal. A sample audio gene generation step of generating an audio gene of an audio sample using the audio gene generation method of claim 1; An audio data audio gene generating step of generating the audio data of the audio data using the audio gene generating method of claim 1; And a search step of searching for audio data including the same audio gene as the sample audio gene. The method of claim 2, The search step, A selection step of selecting a predetermined number of sample audio genes from the sample audio genes; A pseudo value generating step of generating a bit value that is the same as or different from the one sample audio gene; A section searching step of searching for a section having the same value as the similar value among the audio data audio genes, and searching for the audio data audio gene having the same interval as the selected section of the sample audio gene; And calculating whether or not the audio data and the sample audio gene are equal to the audio data including the similar values at equal intervals, and selecting the same audio data as the audio sample only when the difference is equal to or less than a predetermined standard. Audio data retrieval method. The method of claim 3, wherein The audio sample is an audio data retrieval method, characterized in that the amplified above the predetermined signal size from the audio signal input from the wireless communication device and attenuated less than the audio signal transmitted from the wireless communication device from which the noise is removed. The method of claim 4, wherein The audio sample is characterized in that the noise corresponding to a frequency outside the frequency range of the tolerance or the tolerance range of the piano scale is removed. The method of claim 3, wherein The audio sample is an audio data retrieval method, characterized in that the audio signal transmitted directly connected from an audio device such as a radio or television. The method according to claim 1 to 6, In the searching step, the audio genes of the audio data of the database in which the audio data are stored are distributed and temporarily stored in a distributed system, and the search time is shortened by comparing the sample audio genes with the audio genes of the audio data sequentially or simultaneously. Audio data search method characterized in that. The method of claim 7, wherein The searching step further includes a broadcast counting step of calculating the number of broadcasts from the audio device by storing the number of times of audio data selection after the selecting step.