JP2010175908A - Musical piece extraction device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a musical piece extraction device avoiding separation of a musical piece section to the minimum even if wrong clustering is performed temporarily in the middle of the musical piece section. <P>SOLUTION: The musical piece extraction device, which extracts the musical piece section from voice data, includes: a voice data acquisition part which acquires the voice data; a classification part which classifies the acquired voice data portion into a musical piece portions and non-musical piece portions; and an identification part which identifies the start point and the end point of the musical piece section. The identification part detects as the start point a place where the classification result transits from the non-musical piece portion to the musical piece portion, and detects as the end point a place where the result transits from the musical piece portion to the non-musical piece portion. If no musical piece portion is detected in a predetermined period of time after the end point, the end point is defined as the end point of the musical piece section. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a music extraction device that extracts music sections from audio signals.

  An audio signal in a television broadcast or radio broadcast music program or the like generally includes a music (music) section and a non-music section (non-music section). Examples of the non-music section include a section occupied by spoken voice, such as MC [master of ceremony] or DJ [disk jockey].

  In such a situation, some viewers or the like may desire to record only the music section from the audio signal. In this case, it is also possible to record the broadcast and cut out only the favorite music later by editing work.

  However, such editing work is usually cumbersome, and it would be convenient if there was an apparatus that automatically extracted and recorded music sections. In addition, if the music section is automatically extracted and recorded even roughly, it is considered that the burden of the editing work is reduced. For example, according to Patent Document 1, an apparatus for discriminating music and spoken voice using a frequency feature quantity such as MFCC is disclosed.

  The apparatus disclosed in Patent Literature 1 performs classification (clustering) of music or speech for each part (near a cut point) of an audio signal. And the location which changed from speech to music is specified as the start point (start point) of a music section, and the location which changed from music to speech is specified as the end point (end point) of a music section. As a result, a section from the start point to the end point, that is, a section in which portions classified as music pieces continue is extracted as a music section.

JP 2008-241850 A JP 2007-219178 A

  In the music extraction device as described above, it is desirable that clustering is always performed accurately in order to accurately specify the start point and the end point. However, in reality, it is also assumed that erroneous clustering is temporarily performed due to some cause (for example, mixing of noise or the like). In view of this, it is desirable that the music extraction device is designed so that usability is not impaired as much as possible even when erroneous clustering is performed.

  Here, a case is assumed in which erroneous clustering is temporarily performed (in the case of clustering with non-music) in spite of actually being in the middle of a music section. In this case, in the conventional music extraction device, music sections are separated at the places where erroneous clustering is performed. Such a situation also leads to the fact that a music section that should be one in nature is extracted in a state of being separated into a plurality of music sections, which is not preferable from the viewpoint of usability.

  In view of the above-described problems, the present invention provides a music extraction device capable of avoiding the situation where music sections are separated even when temporary clustering is temporarily performed in the middle of the music sections. For the purpose of provision.

  In order to achieve the above object, a music extraction device according to the present invention is a music extraction device that extracts a music section from audio data, and an audio data acquisition unit that acquires audio data and a portion of the acquired audio data as a music A classifying unit that classifies a music piece or a non-music part, and a specifying unit that specifies a start point and an end point of a music section based on the result of the classification, wherein the specifying unit has a result of the classification that is a non-music part If the music part is not detected in the predetermined period after the end point, the end point is detected as the end point. The composition is determined as the end point of the music section.

  According to this configuration, even if clustering (classification) is temporarily performed in the middle of a music section, if correct clustering is performed within the predetermined period, the end point of the music section is incorrect. Setting is prevented. Therefore, it is possible to avoid the situation where the music section is separated as much as possible.

  As described above, according to the music extraction device according to the present invention, even when incorrect clustering (classification) is temporarily performed in the middle of a music section, correct clustering is performed within the predetermined period. This prevents the end point of the music section from being set erroneously. Therefore, it is possible to avoid the situation where the music section is separated as much as possible.

It is a block diagram of the broadcast receiver which concerns on embodiment of this invention. It is a part of flowchart of a music area extraction process. It is a part of flowchart of a music area extraction process. It is explanatory drawing for demonstrating a music area extraction process.

  An embodiment of the present invention will be described below by taking a broadcast receiving apparatus that receives FM radio broadcasts as an example. A block diagram of the broadcast receiving apparatus is shown in FIG. As shown in the figure, the broadcast receiving apparatus 1 includes an FM tuner unit 11, an A / D conversion unit 12, an MP3 codec unit 13, a D / A conversion unit 14, a speaker 15, a DSP 16, a CPU 17, a memory 18, a bus 19, and an HDD 20. And HDD-IF 21 and the like.

  The FM tuner unit 11 performs channel selection processing on FM radio broadcast (analog audio data) continuously input from the preceding stage (for example, an antenna) and transmits it to the subsequent stage. The A / D conversion unit 12 converts the analog audio data continuously transmitted from the FM tuner unit 11 into a digital audio signal (hereinafter referred to as “PCM data”) by a PCM [Pulse Code Modulation] method. Convert and output to the latter stage. This PCM data digitally represents the volume of audio at each location (location where sampling is performed by PCM) in analog audio data.

  In the present application, the term “location” for audio data is a concept representing the position (when the audio appears) when the content of the audio data is represented on the time axis. The sound at different locations in certain sound data is output at different timings when the sound data is reproduced.

  The MP3 Codec unit 13 has a function of compressing PCM data output from the A / D conversion unit 12 by MP3 format encoding and a function of expanding (decoding) the compressed PCM data. For example, PCM data output from the A / D conversion unit 12 is encoded, or PCM data of a music section extracted by the DSP 16 described later is encoded. In addition, the MP3 Codec unit 13 decodes PCM data recorded by compression encoding on the HDD 20.

  The D / A conversion unit 14 converts the PCM data decoded by the MP3 Codec unit 13 into an analog audio signal. The speaker 15 generates sound based on the sound signal transmitted from the D / A conversion unit 14.

  The DSP [Digital Signal Processor] 16 executes various processes according to a program (referred to as an “audio data processing program”) or the like recorded therein in advance. The process includes a process of extracting a section (music section) corresponding to the music from the PCM data and recording it on the HDD 20 (music section extraction process). More detailed contents of the music section extraction process will be described again.

  The CPU [Central Processing Unit] 17 controls various processes executed in the broadcast receiving apparatus 1. The memory 18 temporarily stores audio signals and the like, and stores various information used for processing in the broadcast receiving apparatus 1. Note that the information stored in the memory 18 includes a music model and a non-music model used in the music section extraction process.

  This “music model” is a model that represents the distribution of feature quantities in the frequency domain for some music or for teacher data of music (for example, created by combining representative music of multiple genres). is there. In the present application, the “frequency domain feature amount” may be abbreviated as “frequency feature amount”. This frequency feature amount corresponds to, for example, MFCC [Mel Frequency Cepstral Coefficients]. Further, the “non-music model” is a model that represents the distribution of frequency feature amounts for some non-music (for example, a voice of DJ or the like).

  In addition, the MP3 Codec unit 13, the D / A conversion unit 14, the DSP 16, the CPU 17, and the memory 18 can access each other through a bus 19.

  The HDD [Hard Disk Drive] 20 is a mass storage device, and stores various information such as PCM data encoded in the MP3 format. The HDD 20 is connected to the MP3 Codec unit 13 via an HDD-IF 21 (for example, an ATA interface).

  With the above-described configuration, the broadcast receiving apparatus 1 can execute the above-described music segment extraction process in addition to normal reception of FM radio broadcast. Next, the contents of the music segment extraction process will be described below with reference to the flowcharts shown in FIGS.

  The music segment extraction process is started, for example, when there is an instruction to perform the music segment extraction process by the user. When the music section extraction process is started, the DSP 16 temporarily stores the PCM data continuously output from the A / D converter 12 in the memory 18 one after another (stores until the music section extraction process is completed). Process) is started (step S11). This process is continued until the music segment extraction process is completed.

  In parallel with this process, the following process is first executed to specify the start point of the music section. The DSP 16 checks the amount of change in power (representing the degree of change in the sound level) in order from the earlier location (that is, from the earlier time point) for the PCM data stored in the memory 18. A portion where the amount of change in power is equal to or greater than a predetermined value (Δp) is detected as a first cut point (step S12). That is, a portion where the degree of change in the sound volume in the PCM data is greater than or equal to a predetermined value is detected as the first cut point.

  More specifically, the difference between the power based on the location of interest in the PCM data and the power based on the location before the location of interest is calculated, and when this difference is greater than or equal to Δp, The location to be detected is detected as the first cut point. Note that the calculation of power is realized, for example, by taking the root mean square for one or several samples. The amount of change in power may be, for example, the time differentiation or correlation of the power before and after the above. Further, the process for detecting the subsequent cut points (steps S15, S20, and S24) is also executed by the same method as the process of step S12.

  When the first cut point is detected, the DSP 16 clusters (classifies) the PCM data at the first cut point into music or non-music (step S13). This clustering is realized as follows, for example.

  First, for PCM data, a predetermined area (for example, an area for one second) centered on the first cut point is divided into a predetermined number (for example, 100) parts. Then, the frequency feature amount for each part is calculated. After that, for each part, the likelihood A between the frequency feature quantity and the music model and the likelihood B between the frequency feature quantity and the non-music model are calculated, and any of the likelihood A and the likelihood B is calculated. It is determined whether it is larger.

  For calculating the likelihood, for example, a known method using GMM [Gaussian Mixture Model] is used. Moreover, it can be said that the higher the likelihood between the two, the higher the possibility that they are coincident with each other, or they are more approximate. The likelihood calculation can also be realized by substituting the frequency feature amount for each part into a predetermined evaluation function related to music and non-music registered in the memory 18.

  Then, the ratio of the parts determined that the likelihood A is greater than the likelihood B is calculated as the “approximation degree to the music” at the first cut point. It can be said that the higher the degree of approximation, the higher the possibility that the PCM data at the first cut point belongs to the music section.

  Further, it is determined whether the degree of approximation is equal to or greater than a preset determination threshold Ta. As a result, when it is determined that the degree of approximation is greater than or equal to the determination threshold Ta, the PCM data at the first cut point is clustered into music, and conversely when it is determined that the degree of approximation is less than the determination threshold Ta. The PCM data at the first cut point is clustered into non-music pieces. In addition, the process (step S21 and S25) regarding subsequent clustering is performed by the same method.

  If the PCM data at the first cut point is clustered into non-music as a result of the processing in step S13 described above (N in step S14), it is estimated that the first cut point is not the start point of the music section. Is done. Therefore, the DSP 16 discards the information of the first cut point detected this time, and newly adds a point in the PCM data (after the first cut point detected this time) where the power change amount is equal to or greater than Δp. Are detected as first cut points (step S15). Thereafter, the DSP 16 executes the process of step S13 again.

  On the other hand, when the PCM data at the first cut point is clustered into music (Y in step S14), the DSP 16 sets the first cut point detected this time as the start point of the music section (step S16). . At this time, not the first cut point itself but a predetermined location based on the first cut point (for example, a location separated from the first cut point by a predetermined time) may be set as the start point of the music section. Absent. Thereafter, the following processing is executed in order to specify the end point of the music section.

  The DSP 16 detects, as the second cut point, the next point in the PCM data (after the start point of the music section) where the power change amount is Δp or more (step S20). When the second cut point is detected, the DSP 16 clusters the PCM data at the second cut point into music or non-music (step S21).

  When the PCM data at the second cut point is clustered into music (N in step S22), it is considered that the music section is still continued at the second cut point. Therefore, the DSP 16 discards the information of the second cut point detected this time, and returns to the process of step S20. In the process of step S20 here, the next point in the PCM data (after the second cut point detected this time) where the power change amount becomes Δp or more is newly detected as the second cut point. .

  On the other hand, when the second cut point is clustered into a non-music piece (Y in step S22), it is considered that the music section is basically terminated at the second cut point. However, there is a possibility that, for some reason, the music section is actually continued, but it is temporarily clustered into non-music (that is, erroneous clustering is performed). Therefore, the following processing is further executed in order to more accurately determine whether or not the music section has ended at the second cut point.

  First, the DSP 16 sets the second cut point as a temporary end point (step S23). Thereafter, the next point (after the provisional end point) in the PCM data where the power change amount is Δp or more is detected as a third cut point (step S24). When the third cut point is detected, the DSP 16 clusters the PCM data at the third cut point into music or non-music (step S25).

  When the PCM data at the third cut point is clustered into a song (N in step S26), the PCM data is temporarily clustered into a non-music at the temporary end point, but actually, the song section is still Is presumed to be continued (that is, it should not be the end point of the music section). Therefore, the DSP 16 discards the information on the temporary end point set this time and returns to the process of step S21 (that is, the temporary end point set this time is not regarded as a true end point).

  On the other hand, when the PCM data at the third cut point is clustered into a non-music piece (Y in step S26), the DSP 16 determines whether the interval between the temporary end point and the third cut point is equal to or longer than a predetermined period Ts. Judgment is made (step S27). This period Ts is set in advance as a period (for example, about several seconds) in which the clustering process is considered to be unstable when, for example, temporary noise is mixed in the audio data.

  As a result, if it is less than the period Ts (N in step S27), the process returns to step S24. In the process of step S24 here, the information of the third cut point detected this time is discarded, and the next point (after the third cut point) is a new location where the amount of change in power becomes Δp or more. It will be detected as 3 cut points.

  In the process of step S27, if it is determined that the interval between the temporary end point and the third cut point is equal to or longer than the period Ts (Y in step S27), at least until the period Ts elapses from the temporary end point. The data has never been clustered into music. Therefore, at the temporary end point, the clustering is not temporarily wrong (it is not the fact that it is actually a song but has been clustered into a non-music piece), and the song section may actually end. It can be said that the nature is high.

  The DSP 16 sets the temporary end point as the end point of the music section (step S28). At this time, instead of the temporary end point itself, a predetermined location based on the temporary end point (for example, a location separated from the temporary end point by a predetermined time) may be set as the end point of the music section.

  By the process so far, both the start point and end point of the music section in the PCM data are specified, and as a result, one music section is specified. Therefore, the DSP 16 extracts the music section from the PCM data temporarily recorded in the memory 18. In other words, PCM data representing only the music section is acquired based on the temporarily stored PCM data. Thereafter, the obtained PCM data is subjected to encoding / compression processing in the MP3 Codec unit 13 and then recorded in the HDD 20 as one file (step S29). The broadcast receiving apparatus 1 can reproduce the sound of the recorded PCM data afterwards.

  As a method for extracting a music section, in addition to a technique for acquiring PCM data representing only a music section, the music section is identified by temporarily stored PCM data (PCM data including parts other than the music section). It may be a method of adding information to be enabled (for example, a method of marking the start and end points of a music section).

  Thereafter, the DSP 16 returns to the process of step S12 and repeats the series of processes described above to record the next music section. For example, when the user gives an instruction to end the music segment extraction process, the music segment extraction process is ended.

  Here, in order to make it easier to understand the contents and the like of the music segment extraction process described above, a simple example related to the process will be described below with reference to FIG.

  In this case, it is assumed that the degree of approximation to music in the PCM data temporarily stored in the memory 18 after the start of the music segment extraction processing is as shown in the graph shown in FIG. In the graph of FIG. 4, the horizontal axis indicates time (that is, where in the PCM data), and the vertical axis indicates the degree of approximation to music.

  In addition, each of A to N represents a location where the amount of change in power is equal to or greater than Δp in the PCM data (that is, a location detected as a cut point). A broken line 31 represents the determination threshold Ta described above, and an arrow 32 represents the period Ts described above.

  Note that in the PCM data according to the graph of FIG. 4, in practice, a portion generally indicated by X to a portion indicated by Y is a continuous music section (that is, a single music section). However, in the places indicated by Z1 and Z2, the degree of approximation to the music is temporarily lower than the determination threshold Ta for some reason although it actually belongs to the music section.

  When the music section extraction process is started, the location A is first detected as the first cut point (step S12). However, since the PCM data is clustered into a non-music piece at the location A (N in step S14), the location B is newly detected as the first cut point (step S15).

  However, since the PCM data is clustered into a non-music piece at the location B (N in step S14), the location C is newly detected as the first cut point (step S15). Since the PCM data is clustered into songs at the location C (Y in step S14), the location C is set as the start point of the song section (step S16).

  Thereafter, the portion D is detected as the second cut point (step S20). At the location D, the PCM data is clustered into non-music (Y in step S22), so the location D is set as a temporary end point (step S23). Next, the location E is detected as the third cut point (step S24). However, at the location E, the PCM data is clustered into music (N in step S26), so the temporary end point (the location D) set this time is not set as the end point of the music section. , F are newly detected as second cut points (step S20).

  However, since the PCM data is also clustered into music at the location F (N in step S22), the location G where the power change amount is equal to or greater than Δp is newly detected as the second cut point ( Step S20). Further, since the PCM data is also clustered into music at the location G (N in step S22), the location H at which the power change amount is equal to or greater than Δp is newly detected as the second cut point ( Step S20).

  Since the PCM data is clustered into a non-music piece at the location H (Y in step S22), the location H is set as a temporary end point (step S23). Next, the location I is detected as the third cut point (step S24). However, since the PCM data is clustered into music at the location I (N in step S26), the temporary end point (the location H) set this time is not set as the end point of the music section. In addition, the position of J is newly detected as the second cut point (step S20).

  In addition, since the PCM data is clustered into music at the location J (N at step S22), the location K is newly detected as the second cut point (step S20). Since the PCM data is also clustered into music at the K location (N in step S22), the L location is then newly detected as the second cut point (step S20).

  Since the PCM data is clustered into a non-music piece at the location L (Y in step S22), the location L is set as a temporary end point (step S23). Next, the location M is detected as the third cut point (step S24). At the location M, the PCM data is clustered into a non-music piece (Y in step S26), but as shown in the graph of FIG. 4, the interval between the temporary end point (location L) and the third cut point (location M). Is smaller than the period Ts (N in step S27).

  For this reason, the location N is then detected as the third cut point (step S24), and the PCM data is clustered into music at the location N (Y in step S26). Further, the interval between the temporary end point (L location) and the third cut point (N location) is equal to or longer than the period Ts (Y in step S27). Therefore, the portion L that has been detected as the temporary end point is set as the end point of the music section (step S28). As a result, as indicated by “this embodiment” at the bottom of FIG. 4, the section from C to L in the PCM data is extracted as a single music section and recorded in the HDD 20 (step S29).

  As described above, according to the broadcast receiving apparatus 1, the temporary end point is set as the end point of the music section only when the PCM data is not clustered with the music during the period Ts after the temporary end point. Is done. Therefore, even if erroneous clustering is temporarily performed in the middle of the music section, the music section is not separated at that point (in the above example, “D” or “H”). Yes.

  Note that the PCM data shown in the graph of FIG. 4 is assumed to be based on the conventional method (a method in which a point where a transition from a non-music piece to a music piece is set as the start point of the music section and a point where the music piece is changed to a non-music piece is set as the end point). When the music section is extracted (referred to as “Case 1”), the section from C to D, the section from E to H, and I to L, as shown by “Case 1” at the bottom of FIG. Will be extracted as separate music sections.

  As described above, according to the case 1, one music section is originally recognized by being divided into a plurality of music sections by performing erroneous clustering temporarily. In this regard, it can be said that the broadcast receiving apparatus 1 according to the present embodiment is excellent in usability because it is not recognized separately in this way.

  It should be noted that here, a method in which the same processing as the end point of the music section in the present embodiment is applied to the setting of the start point of the music section (for example, the period from the first cut point at which the clustering result transitions to the music) PCM data shown in the graph of FIG. 4 by a method in which the first cut point is set as the start point of a music section only when there is no clustering with non-music before Ts elapses. Assume that a music segment is extracted (referred to as “Case 2”).

  In this case, since the PCM data is clustered with the non-music at the location D belonging from the location C (the first cut point at which the clustering result transitions to the music) until the period Ts elapses, the location C is It is not set as the start point of the music section. As a result, as indicated by “Case 2” in the lower part of FIG. 4, the location E that is significantly delayed from the starting point of the original music section is set as the starting point of the music section.

  In this respect, in the present embodiment, in setting the start point of the music section, the location where the clustering result has transitioned to the music is set as the start point of the music section, as in the conventional method. That is, when the clustering result transitions to music, the transitioned portion is set as the start point of the music section regardless of the clustering result at the subsequent portion. Therefore, since the location C near the start point of the original music section is set as the start point of the music section, the broadcast receiving device 1 according to the present embodiment can extract the music section more accurately, and more It can be said that it is excellent in usability.

  Thus, in the broadcast receiving apparatus 1, the setting of the start point and the setting of the end point of the music section are realized by processes having different main points. In addition, the broadcast receiving apparatus 1 realizes the setting of the start point and the setting of the end point through processing suitable for each, and improves usability as much as possible.

  In addition, as an aspect of the music section extraction process, each necessary process may be executed in parallel with the reception of the audio data (in real time). It may be executed). Further, for example, a predetermined buffer is provided between the A / D converter 12 and the DSP 16, and PCM data output from the A / D converter 12 is temporarily stored in the buffer. The accumulated PCM data may be transmitted to the DSP 16.

  In this way, instead of performing the process of step S11 in the music section extraction process, the start point of the music section is detected in real time, and the PCM data from the start point is sequentially stored in the memory 18 as soon as the start point is detected. I can let you. It is also possible to detect the end point of the music section in real time so that PCM data after the end point is not stored in the memory 18. That is, it is possible to omit the recording of the part other than the music section in the PCM data in the memory 18. As a result, the capacity of the memory 18 required for temporary storage of PCM data can be suppressed as much as possible.

  In the music segment extraction process, the PCM data clustering is executed only for each cut point (each location where the power change amount is equal to or greater than Δp) (steps S13, S21, and S25). Therefore, it is possible to reduce the processing burden required for the clustering as much as possible. Also, it can be seen that the amount of change in power (degree of change in the amount of sound) in the audio data tends to be relatively large at locations where the audio data transitions from music to non-music or from non-music to music. Yes. Note that the PCM data clustering may be executed at points other than the cut points (for example, uniformly executed at a predetermined cycle).

  As described above, the broadcast receiving apparatus 1 according to the embodiment of the present invention includes a function unit (audio data acquisition unit) that acquires audio data, and whether the acquired audio data part is a music part (part that is a music). A function unit (classification unit) for classifying into a non-music unit (part that is a non-music unit) and a function unit (identification unit) for identifying the start point and end point of a music section based on the result of the classification Yes. And a specific part detects the location which the result of the said classification changed from the non-music part to the music part as a starting point, and detects the part which changed from the music part to the non-music part as an end point (equivalent to the temporary end point mentioned above). . The identifying unit determines the end point as the end point of the music section when the music unit is not detected in a predetermined period after the end point.

  From another point of view, the broadcast receiving apparatus 1 according to the embodiment of the present invention has a function of acquiring audio data, and each location in the audio data (each location where the power change amount is equal to or greater than Δp). A function for classifying music or non-music in order, a function for specifying the start point of the music section in the audio data based on the result of the classification (start point specifying function), and the audio data based on the result of the classification Is provided with a function for specifying the end point of the music section (end point specifying function), and the music section is extracted from the audio data.

  And this end point specifying function uses, as a temporary end point, a location where the result of the classification in the audio data has transitioned to a non-music (that is, a location that has been classified as a music in the immediately preceding location and is classified as a non-music this time). In addition to the detection, a predetermined location based on the temporary end point is specified as the end point only when the audio data is not classified into music in the determination period Ts after the temporary end point.

  For this reason, even if the wrong clustering is temporarily performed in the middle of the music section, if the correct clustering is performed within the predetermined period, the end point of the music section is prevented from being set incorrectly. Is done. Therefore, it is possible to avoid the situation where the music section is separated as much as possible.

  In addition, the start point specifying function detects a location where the result of the classification transitions to music in the audio data (that is, a location that is classified as a non-music at the previous location and is classified as a music this time), and detects the location. A predetermined location as a reference is specified as the starting point. Therefore, it is possible to appropriately specify the start point of the music section as compared to the music extraction device according to case 2 described above.

  Further, the audio data processing program described above is based on the clustering means for classifying the DSP 16 (which can be regarded as a kind of computer) into music or non-music in order from the front for each location in the audio data, Start point specifying means for specifying the start point of the music section in the audio data, end point specifying means for specifying the end point of the music section in the audio data based on the result of the classification, and extraction means for extracting the music section from the audio data; It can be said that the program functions as

  The start point specifying means is a means for detecting a location where the classification result in the audio data has transitioned to music, and specifying a predetermined location based on the location as the start point. In addition, the end point specifying unit detects a portion where the classification result in the audio data has changed to a non-music piece as a temporary end point, and when the audio data is not classified into a music piece in a predetermined period after the temporary end point. As long as it is a means for specifying a predetermined location based on the temporary end point as the end point.

  As mentioned above, although embodiment of this invention was described, this invention is not limited to this content. The present invention can be implemented with various modifications without departing from the spirit of the present invention.

  The present invention can be used in fields such as a receiving apparatus that receives audio data.

1 Broadcast receiving device (music extraction device)
11 FM tuner section 12 A / D conversion section 13 MP3 Codec section 14 D / A conversion section 15 Speaker 16 DSP
17 CPU
18 Memory 19 Bus 20 HDD
21 HDD-IF

Claims (1)

A music extraction device that extracts music sections from audio data,
An audio data acquisition unit for acquiring audio data;
A classification unit for classifying the acquired audio data part into a music part or a non-music part;
Based on the result of the classification, a specifying unit for specifying the start point and end point of the music section;
With
The specific part is:
While detecting the location where the result of the classification has changed from the non-music portion to the music portion as a start point, and detecting the location where the music portion has changed to the non-music portion as an end point,
The music extraction device according to claim 1, wherein when no music part is detected in a predetermined period after the end point, the end point is determined as an end point of the music section.

Images