Audio data processing device and attribute information extraction device

Abstract

To embed attribute information in audio data in a manner that does not affect the quality of the audio data.SOLUTION: An audio data processing device includes an audio data acquisition unit that acquires audio data, an attribute information acquisition unit that acquires attribute information of the audio data, and an audio signal generation unit that generates an audio signal by embedding attribute information in the audio data in such a manner that the data can be decoded in a predetermined non-audible frequency band.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a voice data processing device and an attribute information extraction device. [Background technology]

[0002] 2. Description of the Related Art There is known a technique for embedding power-on or power-off information for turning on or off the power of a mobile terminal as an inaudible signal in the audible band of voice or music. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2006-211624 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, the above-mentioned conventional techniques utilize the audible band, and cannot realize embedding of attribute information into audio data in a manner that does not affect the quality (eg, audibility) of the audio data to which the attribute information is to be embedded.

[0005] Therefore, in one aspect, an object of the present disclosure is to embed attribute information in audio data in a manner that does not affect the quality of the audio data. [Means for solving the problem]

[0006] According to one aspect of the present disclosure, a voice data acquisition unit that acquires voice data; an attribute information acquisition unit for acquiring attribute information of the voice data; and an audio signal generating unit that generates an audio signal by embedding the attribute information in the audio data in a manner that enables the attribute information to be decoded in a predetermined inaudible frequency band. Effect of the Invention

[0007] In one aspect, it becomes possible to embed attribute information in audio data in a manner that does not affect the quality of the audio data. [Brief description of the drawings]

[0008] [Figure 1] 1 is a schematic diagram showing an entire voice data management system according to an embodiment of the present invention; [Diagram 2] 4 is a schematic diagram showing an example of frequency characteristics of audio data and an attribute information signal; FIG. [Diagram 3] FIG. 11 is a schematic block diagram for explaining attribute information embedding processing. [Figure 4] 11 is a diagram showing an example of the timing of embedding attribute information into a time series of voice data. FIG. [Diagram 5] FIG. 11 is a diagram showing an example of catalog information etc. [Figure 6] 4 is a schematic flowchart showing an example of the operation of the audio data processing device according to the present embodiment. [Figure 7] FIG. 2 is a schematic block diagram relating to an extraction process for extracting an attribute information signal from an audio signal. [Figure 8] 11 is a diagram showing an example of a data structure of attribute information in an attribute information signal. FIG. [Figure 9] 1 is a schematic flowchart showing an example of the operation of the attribute information extraction device according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Each embodiment will be described in detail below with reference to the accompanying drawings.

[0010] FIG. 1 is a schematic diagram showing the entirety of a voice data management system 1 according to the present embodiment.

[0011] The voice data management system 1 includes a voice data processing device 10 and an attribute information extraction device 20. Each of the voice data processing device 10 and the attribute information extraction device 20 may be formed by one or more computers. In this case, the computer may be a server computer. Also, the voice data processing device 10 and the attribute information extraction device 20 may be realized by the same computer.

[0012] 1, the voice data processing device 10 and the attribute information extraction device 20 are communicatively connected via a network 5. The network 5 is optional and may include a wireless communication network, the Internet, a Virtual Private Network (VPN), a Wide Area Network (WAN), a wired network, or any combination of these.

[0013] 1 shows an example of functions realized by each of the voice data processing device 10 and the attribute information extraction device 20. Note that some or all of the functions of the voice data processing device 10 described below may be realized by the attribute information extraction device 20, and conversely, some or all of the functions of the attribute information extraction device 20 may be realized by the voice data processing device 10.

[0014] In the following, first, the configuration of the voice data processing device 10 will be described, and then the attribute information extraction device 20 will be described.

[0015] [Audio data processing device] The audio data processing device 10 includes an audio data acquisition unit 102, a filter processing unit 104, an attribute information acquisition unit 106, an audio signal generation unit 108, a catalog information generation unit 110, a first catalog information storage unit 112, a first audio signal data storage unit 114, and an audio signal data transmission unit 116.

[0016] The voice data acquisition unit 102 acquires voice data. The voice data to be acquired is arbitrary, but may have attributes that are suitable for being stored (managed) for a relatively long period of time. For example, the voice data to be acquired may be voice data (data of conversation with a customer) during telephone conversations at a customer center or the like, or voice data during a conference.

[0017] The voice data acquisition unit 102 may acquire voice data in real time, or may acquire voice data from a storage medium (e.g., a voice recorder, etc.). For example, in the example shown in FIG. 1, the voice data acquisition unit 102 acquires voice data in real time via the microphone 2 and the speaker 3. In this case, the voice data acquisition unit 102 acquires human speech data relating to a conversation as voice data via the microphone 2 and the speaker 3. Note that the microphone 2 and the speaker 3 may be provided in, for example, a headset (not shown). Alternatively, the microphone 2 and the speaker 3 may be provided in a mobile terminal or the like. In this case, the voice data acquisition unit 102 may acquire voice data by a wired connection, short-distance communication, etc.

[0018] The filter processing unit 104 executes a predetermined filter process on the voice data acquired by the voice data acquisition unit 102. The predetermined filter process may be a filter process for removing noise or the like. In the present embodiment, as an example, the predetermined filter process includes a process for cutting signals of inaudible frequency components. For example, the filter processing unit 104 may form a bandpass filter that passes only the audible band.

[0019] The attribute information acquisition unit 106 acquires information (hereinafter simply referred to as "attribute information") representing the attributes of the voice data acquired by the voice data acquisition unit 102. The attribute information is arbitrary and may include two or more types of information. The attribute information is acquired for each voice data. Note that a unit of voice data is arbitrary, but in the case of a telephone call, for example, it is one telephone call. Some or all of the attribute information may be generated by a specific user, or may be automatically generated by a program.

[0020] In this embodiment, as an example, the attribute information includes date and time information, a speaker ID, and authority information.

[0021] The date and time information indicates the acquisition date and time of the corresponding voice data. The acquisition date and time of the voice data may be substantially the same as the generation date and time of the voice data. Furthermore, when the voice data acquisition unit 102 acquires voice data in real time, the date and time information may indicate the acquisition date and time of the voice data acquisition unit 102. Note that the granularity of the date and time information is arbitrary, and may be in units of hours or seconds.

[0022] The speaker ID is an identifier representing the speaker related to the corresponding voice data. For example, in the case of answering a telephone call, the speaker ID is the ID of the operator who spoke with the customer. Note that, when there are multiple speakers (speakers excluding the customer), the attribute information related to one voice data may include multiple speaker IDs. Note that each speaker ID may be linked to a name, etc. in a separate database (not shown).

[0023] The authority indicates authority related to the corresponding audio data, and may be, for example, access authority or authority related to editing.

[0024] The attribute information may also include an audio signal ID assigned to the corresponding audio data, information indicating the recording time (data amount) of the corresponding audio data, information indicating the type of telephone answering, conference, etc. The attribute information may also include embedding process information related to the attribute information embedding process (for example, information specifying the predetermined frequency band in which the attribute information is embedded, a predetermined period ΔT, etc.).

[0025] The audio signal generating unit 108 generates a reproducible audio signal based on the audio data acquired by the audio data acquiring unit 102 and the attribute information acquired by the attribute information acquiring unit 106. Specifically, the audio signal generating unit 108 embeds the attribute information acquired by the attribute information acquiring unit 106 into the audio data acquired by the audio data acquiring unit 102 to generate an audio signal. At this time, the audio signal generating unit 108 embeds the attribute information in a manner that makes it decodable in a predetermined inaudible frequency band. More specifically, the audio signal generating unit 108 performs a process of superimposing an electric signal (hereinafter referred to as an "attribute information signal") in which the attribute information is modulated (encoded) using a sine wave of a frequency within a predetermined inaudible frequency band as a carrier wave on the time series of the audio data. For example, a sine wave of a frequency within a predetermined frequency band represents 1 bit, and represents "1" when the signal strength of the attribute information signal exceeds a threshold Th1, and represents "0" otherwise.

[0026] Hereinafter, such processing by the audio signal generating unit 108 is referred to as attribute information embedding processing. According to such attribute information embedding processing, even when the audio signal is reproduced, the attribute information does not affect the reproduction quality (e.g., audibility) of the audio signal. In this way, the audio signal generating unit 108 can embed attribute information in the audio data in a manner that does not affect the quality of the audio data.

[0027] Here, the attribute information embedding process will be further described with reference to Fig. 2 to Fig. 4. Fig. 2 is an explanatory diagram of a predetermined frequency band, with frequency on the horizontal axis and intensity on the vertical axis, showing the frequency characteristics of the voice data and the attribute information signal. For the voice data, Fig. 2 shows the frequency characteristics of the time series s0(t) of the voice data before filtering by the filter processing unit 104, and the frequency characteristics of the time series s1(t) of the voice data after filtering by the filter processing unit 104.

[0028] 2, the attribute information signal has intensity peaks in a plurality of predetermined frequency bands B1, B2, and B3. In this manner, the attribute information signal may be generated in association with two or more predetermined frequency bands.

[0029] 3 is a block diagram showing an outline of the attribute information embedding process related to the attribute information signals m1(t), m2(t), etc. In this embodiment, the attribute information signal generating units 1081, 1082, etc. of the audio signal generating unit 108 generate attribute information signals m1(t), m2(t), etc. corresponding to the predetermined frequency bands B1, B2, etc. In the example shown in FIG. 3, a time series s1(t) after filtering in the audible band A1 is generated from a time series s0(t) of original audio data, and two or more attribute information signals m1(t), m2(t), etc. are superimposed on the time series s1(t) after filtering, thereby generating an audio signal s2(t).

[0030] FIG. 4 shows an example of the timing of embedding attribute information into a time series s0(t) of audio data. In FIG. 4, the time period (timing) at which attribute information is embedded is shown typically as a time series range R1. As shown in FIG. 4, embedding of attribute information may be performed multiple times throughout the entire time series of audio data. In this case, the attribute information may be embedded at predetermined intervals ΔT as shown in FIG. 4, or may be embedded irregularly. Alternatively, in another embodiment, the attribute information may be embedded only in a certain time period of the time series of audio data (for example, a predetermined time after the start point).

[0031] When the audio signal generating unit 108 generates an audio signal, which is audio data in which attribute information is embedded, in this manner, the audio signal generating unit 108 stores the audio signal in the first audio signal data storage unit 114.

[0032] The catalog information generating unit 110 generates catalog information regarding the voice data acquired by the voice data acquiring unit 102. The catalog information is information that can associate the voice data acquired by the voice data acquiring unit 102 with the attribute information acquired by the attribute information acquiring unit 106.

[0033] Fig. 5 shows an example of catalog information 500. In the example shown in Fig. 5, catalog information 500 includes various types of other attribute information in association with an audio signal ID given to each piece of audio data.

[0034] The first catalog information storage unit 112 stores the catalog information generated by the catalog information generation unit 110. Each time voice data is acquired, new catalog information corresponding to the acquired new voice data may be accumulated in the first catalog information storage unit 112. The catalog information data in the first catalog information storage unit 112 may be used to extract a specific voice signal from the first voice signal data storage unit 114. For example, a user of the voice data management system 1 may be able to extract a desired data portion from the catalog information data by specifying a period, a speaker ID, etc., and display and play the data on his / her own user terminal (e.g., a smartphone or a PC).

[0035] The first audio signal data storage unit 114 stores the audio signal generated by the audio signal generation unit 108. Each time audio data is acquired, a new audio signal corresponding to the acquired new audio data may be accumulated in the first audio signal data storage unit 114. The audio signal stored in the first audio signal data storage unit 114 may be linked to an audio signal ID, for example, as in the audio signal data information 600 shown in FIG. 5, and may be extractable using the audio signal ID as a keyword. Therefore, when a user of the audio data management system 1 identifies an audio signal ID related to a desired audio signal, for example, by referring to catalog information, the user can extract the desired audio signal from the first audio signal data storage unit 114 based on the identified audio signal ID.

[0036] The voice signal data transmission unit 116 transmits a part or all of the voice signal data in the first voice signal data storage unit 114 to the attribute information extraction device 20.

[0037] However, system migration related to the voice data management system 1 tends to require a huge amount of work for checking, etc., since it handles a large amount of voice data accumulated over a relatively long period of time. In addition, the link between the voice data and attribute information is easily broken, and it is difficult to prevent voice data replacement or misplacement.

[0038] In this respect, in this embodiment, as described above, the voice data is managed / stored as a voice signal with attribute information embedded therein, so that system migration becomes easy. For example, even if a part or all of the catalog information data in the first catalog information storage unit 112 is destroyed, after the system migration, the voice data can still be managed based on the attribute information embedded in the voice signal.

[0039] In the present embodiment, the filter processing unit 104 is provided, but the filter processing unit 104 may be omitted. In this case, noise in an inaudible frequency band may be included in the audio data acquired by the audio data processing device 10. If significant noise is mixed into the predetermined frequency band in which the attribute information is embedded, it may be difficult to extract the attribute information.

[0040] Therefore, in this embodiment, a frequency band determination unit 105 may be provided. In this case, the frequency band determination unit 105 determines a predetermined frequency band from among frequency bands in which the signal strength is lower than the threshold Th2 based on the relationship between the frequency and the signal strength obtained by frequency analysis of the voice data acquired by the voice data acquisition unit 102. Specifically, the frequency band determination unit 105 identifies a frequency band in which the noise level is lower than the threshold Th2 among inaudible frequency bands, and determines a predetermined frequency band from among the frequency bands. This makes it possible to realize the attribute information embedding process in a manner that is less susceptible to the influence of noise even in an environment in which noise may be mixed into the voice data. The threshold Th2 may be the same as the above-mentioned threshold Th1, or may be significantly smaller than the above-mentioned threshold Th1.

[0041] Next, an example of the operation of the audio data processing device 10 according to this embodiment will be described with reference to Fig. 6. Fig. 6 is a schematic flow chart showing an example of the operation of the audio data processing device 10 according to this embodiment. The process shown in Fig. 6 may be executed each time a piece of audio data is acquired by the audio data acquisition unit 102. Alternatively, when a plurality of pieces of audio data are acquired collectively by the audio data acquisition unit 102, the process shown in Fig. 6 may be executed for each piece of audio data.

[0042] In step S602, the audio data processing device 10 judges whether or not the data acquisition flag F1 = 0. When the data acquisition flag F1 is "0", it indicates that no audio data has been acquired by the audio data acquisition unit 102, and when the data acquisition flag F1 is "1", it indicates that audio data has been acquired by the audio data acquisition unit 102. If the judgment result is "YES", the process proceeds to step S604, and if the judgment result is "NO", the process proceeds to step S614.

[0043] In step S604, the audio data processing device 10 judges whether or not a new piece of audio data has started to be acquired by the audio data acquisition unit 102. If the judgment result is "YES", the process proceeds to step S606, otherwise the process goes into a standby state in which the process waits for acquisition of a new piece of audio data.

[0044] In step S606, the audio data processing device 10 assigns an audio signal ID to the currently acquired piece of audio data. This audio signal ID is associated with an audio signal (see step S620) stored in the first audio signal data storage unit 114 based on the currently acquired piece of audio data.

[0045] In step S608, the audio data processing device 10 starts storing in the first audio signal data storage unit 114 the one piece of audio data that has just started to be acquired.

[0046] In step S610, the audio data processing device 10 sets the data acquisition flag F1 to "1".

[0047] In step S612, the voice data processing device 10 acquires (generates) attribute information relating to the current piece of voice data, and the processing for the current cycle ends.

[0048] In step S614, the audio data processing device 10 judges whether or not the timing for embedding attribute information (see FIG. 4) has arrived. When the timing for embedding attribute information is set for each predetermined period ΔT (see FIG. 4), it may be judged whether or not the timing for embedding attribute information has arrived based on a timer that times out for each predetermined period ΔT. If the judgment result is "YES", the process proceeds to step S616, and otherwise the process proceeds to step S617.

[0049] In step S616, the voice data processing device 10 embeds the attribute information generated in step S612 into the current one voice data. That is, the voice data processing device 10 executes the attribute information embedding process described above. In this case, the currently acquired part of the currently acquired one voice data is stored in the first voice signal data storage unit 114 with the attribute information embedded therein.

[0050] In step S617, the audio data processing device 10 stores the currently acquired piece of audio data for the current period in the first audio signal data storage unit 114 in its original form (that is, in a form in which no attribute information is embedded).

[0051] In step S618, the audio data processing device 10 judges whether or not the acquisition of the current set of audio data has been completed. If the judgment result is "YES", the process proceeds to step S620, otherwise the process of the current cycle ends.

[0052] In step S620, the audio data processing device 10 generates catalog information related to the current one audio data, and stores it in the first catalog information storage unit 112. Note that step S620 may be executed at an earlier stage (i.e., before the determination result in step S618 becomes YES).

[0053] In step S622, the audio data processing device 10 resets the data acquisition flag F1 to "0".

[0054] 6, the audio signal data in the first audio signal data storage unit 114 can be updated for each piece of audio data based on the audio signal obtained by embedding attribute information in real time while acquiring the audio data. Also, catalog information can be generated for each piece of audio data, and the catalog information data in the first catalog information storage unit 112 can be updated.

[0055] [Attribute information extraction device] During system migration, recovery from damage to catalog information data in first catalog information storage unit 112, etc., attribute information extraction device 20 extracts attribute information embedded in the audio signal based on the audio signal in first audio signal data storage unit 114. Then, attribute information extraction device 20 associates the extracted attribute information with the audio signal from which it was extracted, thereby enabling various data management for system migration, such as new creation of catalog information, etc.

[0056] In this embodiment, the attribute information extraction device 20 includes a signal acquisition unit 202 , an attribute information extraction unit 204 , a linking processing unit 206 , a second catalog information storage unit 208 , and a second audio signal data storage unit 210 .

[0057] The signal acquiring unit 202 acquires a voice signal generated by the voice data processing device 10. Specifically, the signal acquiring unit 202 acquires a voice signal transmitted from the voice signal data transmitting unit 116 of the voice data processing device 10. For example, a user of the voice data management system 1 requests a part or all of the voice signal data in the first voice signal data storage unit 114 when migrating the system. In this case, the signal acquiring unit 202 may transmit request information to the voice data processing device 10, and the voice data processing device 10 may transmit the requested voice signal to the attribute information extraction device 20 in response to the request information.

[0058] The attribute information extraction unit 204 extracts attribute information based on the audio signal acquired by the signal acquisition unit 202 .

[0059] Here, an example of a method for extracting attribute information by the attribute information extraction unit 204 will be described with reference to FIG. 7 and FIG.

[0060] 7 is a block diagram showing an extraction process for extracting attribute information signals m1(t), m2(t), ..., which are generated in association with a plurality of predetermined frequency bands, from a voice signal. In the example shown in FIG. 7, two or more attribute information signals m1(t), m2(t), ... are extracted by performing a filter process on the voice signal s2(t) using the filter units 2041, 2042, ..., of the attribute information extraction unit 204. In this case, the filter units 2041, 2042, ..., are configured as band-pass filters that pass only signals of frequency components of the predetermined frequency bands B1, B2, .... As a result, the attribute information signals m1(t), m2(t), ..., can be extracted via the filter units 2041, 2042, ....

[0061] 7, it is assumed that the plurality of predetermined frequency bands in which the attribute information signals m1(t), m2(t), ... are embedded do not change for each audio signal s2(t) and are constant, but this is not limiting. For example, when the above-mentioned frequency band determination unit 105 is provided in the audio data processing device 10, the functioning filter unit among the filter units 2041, 2042, ... may differ depending on the predetermined frequency band determined by the frequency band determination unit 105.

[0062] Fig. 8 shows an example of a data structure 80 of attribute information in an attribute information signal. As shown in Fig. 8, the data structure 80 of attribute information in an attribute information signal may include a header portion 800 and a data portion 801. The data portion 801 includes attribute information. In this case, when the attribute information extraction unit 204 detects the header portion 800 of each of the attribute information signals m1(t), m2(t), ..., it can extract (decode) the attribute information from the data portion 801.

[0063] According to this embodiment, as described above, the attribute information signals m1(t), m2(t), ... are subjected to the attribute information embedding process in different predetermined frequency bands, so that even if some of the attribute information signals m1(t), m2(t), ... cannot be extracted, the attribute information can be extracted based on at least one of the plurality of predetermined frequency bands. In other words, by increasing redundancy, the attribute information can be reliably extracted.

[0064] Furthermore, according to this embodiment, as described above, for one time series of voice data, the attribute information signals m1(t), m2(t), ... include attribute information in multiple time periods (see Fig. 4), so even if attribute information cannot be extracted in some time periods, it is possible to extract attribute information in other time periods. In other words, by increasing redundancy, it is possible to reliably extract attribute information.

[0065] Incidentally, in this embodiment, since the attribute information signals m1(t), m2(t), ... contain the same attribute information, it is possible to extract the same attribute information from each of the attribute information signals m1(t), m2(t), ...

[0066] However, due to disturbances such as noise, some of the attribute information extracted from each of the attribute information signals m1(t), m2(t), ... may not be consistent. For example, the attribute information extracted from the attribute information signal m1(t) and the attribute information extracted from the attribute information signal m2(t) may differ from each other. In such a case, for example, when there are three or more types of predetermined frequency bands, the candidate with the highest consistency among the candidates of attribute information extracted from the attribute information signals of each predetermined frequency band may be determined as the final attribute information based on the principle of majority rule. Alternatively, when attribute information is embedded in a plurality of time periods in one voice data, the candidate with the highest consistency among the candidates of attribute information extracted in each time period may be determined as the final attribute information based on the principle of majority rule.

[0067] The linking processing unit 206 links the attribute information extracted by the attribute information extraction unit 204 with the audio signal from which it was extracted. That is, the linking processing unit 206 generates new catalog information (e.g., catalog information after system migration). The new catalog information may be the same as the original catalog information (catalog information in the first catalog information storage unit 112), or may be modified in various ways. For example, a new ID may be assigned to the audio signal ID. The new catalog information may include, as new attribute information, information indicating that the new attribute information has been generated based on the attribute information extracted by the attribute information extraction unit 204, and may include date and time information.

[0068] The second catalog information storage unit 208 stores new catalog information (e.g., catalog information after system migration) generated by the linking processing unit 206. Note that the second catalog information storage unit 208 may have a function similar to that of the first catalog information storage unit 112 of the voice data processing device 10, for example, after system migration. For example, when new voice data is acquired by the voice data acquisition unit 102 after system migration, catalog information related to the voice data may be stored in the second catalog information storage unit 208.

[0069] The second audio signal data storage unit 210 stores the audio signal acquired by the signal acquisition unit 202. Note that the second audio signal data storage unit 210 may have a function similar to that of the first audio signal data storage unit 114 of the audio data processing device 10, for example, after system migration. For example, when audio data is newly acquired by the audio data acquisition unit 102 after system migration, the audio signal related to the audio data may be stored in the first audio signal data storage unit 114.

[0070] Next, an operation example of the attribute information extraction device 20 according to this embodiment will be described with reference to Fig. 9. Fig. 9 is a schematic flowchart showing an operation example of the attribute information extraction device 20 according to this embodiment. The process shown in Fig. 9 may be executed every time one voice signal is acquired by the signal acquisition unit 202. Alternatively, if the voice signal data in the first voice signal data storage unit 114 has already been acquired by the signal acquisition unit 202 and stored in the second voice signal data storage unit 210, the process shown in Fig. 9 may be executed every time one voice signal is read out from the second voice signal data storage unit 210. Note that one unit of the voice signal corresponds to one unit of voice data.

[0071] In step S 902 , the attribute information extraction device 20 stores the voice signal acquired from the voice data processing device 10 in the second voice signal data storage unit 210 .

[0072] In step S904, the attribute information extraction device 20 sets a value i representing a processing order number to i=1.

[0073] In step S906, the attribute information extraction device 20 starts reproducing the attribute information signal m1 embedded in the i-th predetermined frequency band among a plurality of predetermined frequency bands (here, N predetermined frequency bands). That is, the attribute information extraction device 20 starts reproducing the attribute information signal m1 output from a filter unit that passes only the components of the i-th predetermined frequency band among the filter units 2041, 2042, ....

[0074] In step S908, the attribute information extraction device 20 judges whether or not the header portion 800 (see FIG. 8) is detected in the signal portion cut out by the window function. The header portion 800 may be detected based on the time when the signal strength of the i-th predetermined frequency band exceeds the threshold value Th1, for example. If the judgment result is "YES", the process proceeds to step S910, and otherwise the process proceeds to step S909.

[0075] In step S909, the attribute information extraction device 20 extracts a new signal portion using a window function and calculates a power spectrum (i.e., performs FFT analysis). In this way, using the window function, the reproduction of the audio signal (changing the extracted portion) is continued until the header portion 800 is detected. Note that, if the time period in which the attribute information is embedded is known, the window function may be applied only to the portion of the attribute information signal m1 that corresponds to the known time period.

[0076] In step S910, the attribute information extraction device 20 records time point information (time-series position information within the audio signal) for the header portion 800 detected this time.

[0077] In step S912, the attribute information extraction device 20 extracts attribute information (candidates) based on the data portion 801 following the currently detected header portion 800.

[0078] In step S914, the attribute information extraction device 20 judges whether or not the reproduction of the attribute information signal m1 embedded in the i-th predetermined frequency band is completed. Note that, in the case where the attribute information is embedded in only one time period in each of the multiple predetermined frequency bands, step S914 may be omitted. If the judgment result is "YES", the process proceeds to step S916, otherwise, the process from step S908 is repeated via step S909.

[0079] In step S916, the attribute information extraction device 20 increments the value i representing the processing order number by "1".

[0080] In step S918, the attribute information extraction device 20 judges whether i=N. That is, the attribute information extraction device 20 judges whether the extraction process for the attribute information signal m1 related to each of the plurality of predetermined frequency bands is completed. If the judgment result is "YES", the process proceeds to step S922, otherwise, the process from step S906 is repeated.

[0081] In step S922, based on the result of recording the time point information in step S910, it is determined whether or not there are header portions 800 detected in the same time period in each of the multiple predetermined frequency bands. In other words, it is determined whether or not the data portions 801 start from the same time point. Here, Rn(i) is assumed to be the natural noise at time x of the i-th predetermined frequency band fi. Here, N=3, and R1 to R3 are assumed to have no correlation. In this case, the signal strengths (audio levels) L1(x), L2(x), and L3(x) of the attribute information signals m1(t), m2(t), and m3(t) at time x (i.e., t=x) can be expressed as follows: L1(x)=P1(x)+R1(x) L2(x)=P2(x)+R2(x) L3(x)=P3(x)+R3(x) P1(x), P2(x), and P3(x) are the signal intensities of the attribute information signals m1(t), m2(t), and m3(t), respectively, and are the signal intensities of signal components other than noise. Therefore, when the time point x is in the time period corresponding to the header section 800, P1(x), P2(x), and P3(x) are greater than the threshold value Th1. On the other hand, when the time point x is outside the time period corresponding to the header section 800, it is unlikely that all of P1(x), P2(x), and P3(x) are greater than the threshold value Th1. From such characteristics, it is found that the data section 801 (attribute information) following the header section 800 detected in the same time period is less likely to be a data section that may be erroneously detected due to noise or the like, and therefore the attribute information can be extracted with high accuracy.

[0082] Specifically, in step S922, if the determination result is "YES", the process proceeds to step S924, otherwise the process proceeds to step S926.

[0083] In step S924, the attribute information extraction device 20 extracts attribute information based on the header portion 800 detected in the same time period (extracts it as the final attribute information). That is, the attribute information is extracted based on the data portion 801 that starts from the same time point (extracts it as the final attribute information).

[0084] In step S926, the attribute information extraction device 20 extracts attribute information based on the header portion 800 detected in another time period (extracts it as final attribute information). For example, a time period in which the number of attribute information signals in which the header portion 800 is detected in the same time period is identified, and attribute information is extracted based on the header portion 800 detected in that time period (extracts it as final attribute information).

[0085] In step S928, the attribute information extraction device 20 generates new catalog information based on the attribute information extracted in step S924 or step S926, and stores the new catalog information in the second catalog information storage unit 208.

[0086] 9, the attribute information can be extracted (decoded) based on the audio signal in which the attribute information is embedded in the audio data, and new catalog information can be generated. Therefore, even if the catalog information in the first catalog information storage unit 112 is partially or entirely destroyed, the catalog information can be restored.

[0087] 9, in particular, a plurality of attribute information signals are used to determine the consistency of the time information at which the header portion 800 is detected (see step S922), thereby reducing the possibility that the header portion (and the data portion associated therewith) will be erroneously detected due to noise, etc. This allows the attribute information to be extracted with high accuracy.

[0088] According to the present embodiment described above, the following effects are particularly achieved.

[0089] If attribute information is managed separately from the voice data, there are problems such as the need to transfer the voice data and attribute information in small chunks in order to maintain consistency between the data and the data, which is time-consuming, the difficulty of checking the data after transfer, and the difficulty of discovering transfer errors.

[0090] In contrast, according to this embodiment, by incorporating attribute information into the voice data and extracting it at the time of transfer, it is easy to maintain the consistency between the voice data and the attribute information, the labor and confirmation work for transfer can be reduced, and mistakes at the time of transfer are less likely to occur. Also, by incorporating attribute information in the inaudible area, the voice data can be played back on a normal audio player, and there is an effect that it does not affect the playback and confirmation of recorded data in call center operations, for example.

[0091] Although each embodiment has been described above in detail, the present invention is not limited to the specific embodiment, and various modifications and changes are possible within the scope of the claims. In addition, it is also possible to combine all or a plurality of the components of the above-described embodiments. [Explanation of symbols]

[0092] 1. Voice Data Management System 2. Microphone 3 Speakers 5. Network 10 Audio data processing device 20 Attribute information extraction device 102 Voice data acquisition unit 104 Filter processing section 105 Frequency band determination unit 106 Attribute information acquisition section 108 Audio signal generator 110 Catalog Information Generation Unit 112 First catalog information storage unit 114 First audio signal data storage unit 116 Audio signal data transmission unit 202 Signal Acquisition Unit 204 Attribute information extraction section 206 Linking Processing Unit 208 Second catalog information storage unit 210 Second audio signal data storage unit

Claims

1. A signal acquisition unit that acquires an audio signal; an attribute information extraction unit that extracts attribute information embedded in the audio signal in a manner that enables decoding in a predetermined inaudible frequency band from the audio signal acquired by the signal acquisition unit; a linking processing unit that links the attribute information extracted by the attribute information extraction unit with the audio signal; An attribute information extraction device comprising:

2. the audio signal is generated by an audio data processing device; The audio data processing device includes: a voice data acquisition unit that acquires voice data; an attribute information acquisition unit for acquiring attribute information of the voice data; The attribute information extraction device according to claim 1 , further comprising: a voice signal generation unit that generates a voice signal by embedding the attribute information in the voice data in the manner.

3. The attribute information extraction device according to claim 2 , wherein the audio signal generation unit embeds the attribute information in each of a plurality of the predetermined frequency bands.

4. An attribute information extraction device as described in claim 2 or 3, wherein the audio signal generation unit embeds the attribute information multiple times within a certain period of time of the time series of the audio data, or throughout the entire time series of the audio data.

5. The audio data processing device includes: The attribute information extraction device according to any one of claims 2 to 4, further comprising a frequency band determination unit that determines the predetermined frequency band from among frequency bands in which the signal strength is lower than a threshold value based on a relationship between a frequency obtained by frequency analysis of the voice data and a signal strength.

6. The audio data includes human speech data, 5. The attribute information extraction device according to claim 2, wherein the audio signal generation unit generates the audio signal in real time in parallel with recording of the person's speech.

7. The attribute information extraction device described in Claim 3, wherein the attribute information extraction unit extracts candidates for the attribute information in each of the multiple specified frequency bands, and extracts the attribute information based on the extracted multiple candidates for the attribute information.

8. The attribute information extraction device of Claim 4, wherein the attribute information extraction unit extracts the attribute information based on a signal portion of the time series of the audio signal that relates to a specified time period or an arbitrary time period.

9. An attribute information extraction device as described in Claim 5, wherein the attribute information extraction unit extracts the attribute information in the specified frequency band determined by the frequency band determination unit of the audio data processing device.

Images