JP2002304616A - Music selection support method, program, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a music selection support method which can easily select music matching mental and physical states. SOLUTION: State vectors are computed from pulse wave data (steps 202 to 224), sensitive vectors are obtained by a neural network (steps 226 to 230), and the differences between the initial sensitive vector and respective sensitive vectors are computed to select the best music (steps 232, 234).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a music selection support method, a program, and a recording medium, and more particularly to a music selection support method and a music selection support method for supporting selection of a desired music from a large number of music. The present invention relates to a computer-executable program and a recording medium.

[0002]

2. Description of the Related Art At present, music lovers can search for songs advertised on their favorite singers and televisions using a music selection support system or the like on a music distribution site and download them to a personal computer, a mobile phone, or a PHS. It is possible. Such a music distribution site distributes music requested by the user in a compressed file format such as ATRAC3 via the Internet or using a mobile phone network. A music selection support system is disclosed in, for example, Japanese Patent Application Laid-Open No. 2000-33090 in which music selection is supported by narrowing down various types of information (for example, female solo cigars, dynamism, etc.) that can be expressed in characters in relation to the music itself. Known techniques are known.

[0003] People's perception of music is often expressed in words and letters, but as essential and objective information,
Research using biological signals such as brain waves and pulse waves has been conducted in various fields. For example, the sound of insects may sound noise to one person and relieve another. Examining the function of the brain, it is known that in the former case, the left brain is working, while in the latter case, the right brain is working. Techniques for objectively diagnosing the physical and mental state of a person include, for example, JP-A-6-54813 and JP-B-6-954.
There is a technique using a chaos attractor and a Lyapunov exponent described in Japanese Patent No.

[0004] By the way, it is said that many of the causes of semi-health in urban Saran-Man are due to stress prevailing in modern society. Not only adults but also young people are under stress, and many diseases are caused by it. In addition, the stress state deprives free thinking and accurate judgment. On the other hand, it is known that the onset of Alzheimer's disease can be delayed by listening to music that is comfortable for an individual. It is said that if the onset of Alzheimer's disease can be delayed for five years, the medical cost for healing dementia will be halved, which is significant.

[0005]

SUMMARY OF THE INVENTION However, Japanese Patent Application Laid-Open
According to the technology disclosed in Japanese Patent Application Laid-Open No. 000-33090, a desired music can be selected from a huge number of music of various genres from past to latest, but the music selected by himself can be relaxed by himself / herself. It is not always optimal. In other words, how people feel when listening to music
It depends greatly on the person's past life experience, current social environment, psychological condition, and the external environment when listening to the music (indoor temperature, whether on a train, outside, etc.) Generally, reputable music is not necessarily optimal for all users if it gives comfort.
The music that the same person feels comfortable on the day may not be so much the next day. Conversely, there are cases in which the subconscious mind is not relaxed even if they think they are relaxing, and an objective evaluation that accurately reflects mental and physical information is required.

If the music selection work for relaxing the user is not easily performed in real time at each timing, the act of selecting comfortable music itself may be stressful. Until now,
As in the technology of 54813 and Japanese Patent Publication No. 6-9546, studies of measuring respiration, heart rate, brain waves, etc., performing frequency analysis and the like, and evaluating the state of the heart have been conducted in various fields. Since the device is expensive, it has been difficult to adopt it for general household use.

The present invention has been made in view of the above circumstances, and provides a music selection support method capable of easily selecting music suitable for a physical and mental state, a program executable by the music selection support method, and a recording medium on which the program is recorded. As an issue.

[0008]

In order to solve the above-mentioned problems, a first aspect of the present invention is a music selection support method for supporting selection of a desired music from a large number of musics. A state vector representing the characteristic amount of the chaos attractor is calculated from the pulse wave time series data before listening to the music and the plurality of pulse wave time series data when listening to a plurality of music, and the state vector is input to the neural network. Input to the layer, and acquire a sensitivity vector representing a feature amount of sensitivity from the output layer for each of the pulse wave time series data before listening to the music and the plurality of pulse wave time series data when listening to a plurality of music. Then, the increase / decrease tendency of the difference between the sensitivity vector of the pulse wave time series data before listening to the music and the sensitivity vector of the pulse wave time series data when listening to the music is calculated, respectively. Selecting a song having an increasing / decreasing tendency that matches the increasing / decreasing tendency of the sensitivity vector set in advance corresponding to the song as a song from which the subject can obtain a desired emotional state, among the increasing / decreasing trends of the difference. .

In this case, the elements of the state vector preferably include at least a Lyapunov exponent and a correlation dimension, and the sensitivity vector includes at least stress,
It is even more preferred to include a sensation element of sadness, joy and relaxation. At this time, receiving the pulse wave time-series data before listening to the music of the subject from the moving object and a plurality of pulse wave time-series data when listening to a plurality of music, and transmitting the selected music to the moving object You may make it.

A second aspect of the present invention is a computer-executable program for executing the first aspect, and a third aspect of the present invention is a computer-readable recording medium storing the program.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of a music selection service to which the present invention can be applied will be described with reference to the drawings.

(Configuration) First, the schematic configuration of each site where the music selection service of the present embodiment is executed will be described.

As shown in FIG. 1, the Internet 5 includes a packet center 3 for connecting packet communication with the mobile phone 1 via the ground relay station 4 to the Internet 5. A music selection site 6 for selecting the most suitable music to bring the user to a desired psychological state, such as divergence, and notifying the user of the music, a music distribution site for allowing the user to listen to the sample music in a streaming format and delivering the music desired by the user. Sites 7a, 7b,
7c (hereinafter, these are collectively referred to as a music distribution site 7). These music distribution sites 7
Is operated by a so-called record company.

The mobile phone 1 has a display for displaying an image, a web page, and the like, a cursor direction button for moving a cursor displayed on the display in the up, down, left and right directions. A confirm button and a numeric keypad for confirming the input information are arranged. The mobile phone 1 has a telephone function,
In addition to the function of connecting to the Internet, it has a function of downloading a JAVA (registered trademark) application from various servers on the Internet and executing it in the mobile phone. A built-in SD memory can record data. A UIM (User Identity Module) in which information on a telephone number and a memory dial is recorded may be used as the memory. In addition, a short distance (about 10
Bluetooth for establishing an ad-hoc (can be connected only when needed, not always) wireless connection between the mobile phone 1 and the pulse wave measuring device 2 for measuring a user's pulse wave within m)
h module (hereinafter, abbreviated as BTM) is incorporated.

[0015] Bluetooth is a short-range wireless link intended to replace a cable between a mobile object and a fixed electronic device (for example, a pulse wave meter 2), and was proposed by Ericsson in 1994 and is currently a Japanese company. 2000 including
Has more than one sponsoring company. BTM is Bluet
The baseband section that defines the basic functions of the ooth and the RF section that generates a high-frequency radio frequency that is a protocol layer are housed in one chip, and have an interface section that allows connection with external devices such as fixed electronic devices. are doing.

As shown in FIG. 2, a pulse wave measuring device 2 comprises a pulse wave sensor 21 for detecting a pulse wave of a finger of a user,
1, an A / D converter 23 for converting an analog signal output from the operational amplifier 22 into a digital signal at a sampling rate of 1,000 Hz, and a digital signal output from the A / D converter 23. BTM 24 for establishing an ad-hoc wireless link with the BTM in the mobile phone 1, a small antenna 26 for diffusing the high frequency output from the BTM 24 into the air, and a power supply unit 25 for converting a commercial power supply and supplying a constant voltage to each unit. It is composed of

As shown in FIG. 3A, the pulse wave sensor 21
Has a cover portion 42 made of a black cylindrical sponge rubber with a flexible and light-shielding property, into which a fingertip 41 of a user (subject) can be inserted. The lower part of the cross section of the cover part 42 is thicker than the upper part. (See also FIG. 3B). The infrared light emitting diode 45 and the phototransistor 46 are provided with a convex lens at the tip, and the optical axis 45a of the infrared light emitting diode 45 is located at a position where the tip of the biconvex lens can contact the abdominal side 43 of the fingertip 41.
And the optical axis 46a of the phototransistor 46 is 20 ° to 3 °.
It is buried so as to intersect at an angle of 0 °. Therefore,
Since the pulse wave sensor 21 is formed of a flexible and light-blocking material in the shape of a bottomed cylinder, the fingertip 41 can be easily attached and detached, and the detection of the pulse wave is not affected by external light and the infrared light emitting diode 45 and the phototransistor 46 can be stably (fixedly) mounted.

As shown in FIG. 3B, the anode of the infrared light emitting diode 45 is connected to the Vcc side of the power supply 25 via the resistor R1, and the cathode is connected to the ground (G) of the power supply 25. Have been. The collector of the phototransistor 46 is connected to the Vcc side via the resistor R2, and the emitter is connected to the ground side. An output terminal OUT for outputting a voltage signal is derived from a connection point between the collector and the resistor R2, and this output terminal OUT is connected to the input side of the operational amplifier 22.
Therefore, the pulse wave sensor 21 detects the infrared light emitting diode 45
Infrared light from the user is reflected inside the fingertip 41, received by the phototransistor 46, and the pulse wave of the user can be output as a voltage.

As shown in FIG. 4, the music selection site 6
An armed zone 61 such as a firewall for armed against unauthorized access from an external network is provided. The Internet 5 is connected to a bus-type LAN 62 set up in the music selection site 6 via the armed zone 61. The LAN 62 includes a music selection server 65 for notifying the user of the URL of the music distribution site 7 having the most suitable music for bringing the member to a desired psychological state, member data such as member name and age, music attribute data, and the like. A database server 66 storing the URL and the like of the music distribution site 7,
An analysis server 67 for selecting and analyzing the most suitable music for the user from a plurality of songs based on the user's current psychological state and the psychological state after sample listening of the sample music, a web server 64 for storing a homepage and performing member registration, Membership fee management that manages information for adding a flat fee to the member and charging the member when the member is billed for the monthly flat fee or when the music distribution site 7 notifies the member of the charged amount as described later. The server 63 is connected.

As shown in FIG. 5, the music distribution site 7a has an armed zone 71, and the Internet 5 is connected to a LAN 72 provided in the music distribution site 7a via the armed zone 71. The LAN 72 stores a database such as a correspondence table between a song name and an IP address of a music server and stores various music data in a streaming format or a download format in accordance with the received contents. A music distribution server 73 for transmitting data and a charging server 76 for calculating a charging amount according to the number of songs downloaded (distributed) to the member when full music data is transmitted in a download format are connected. The music distribution site 7
b and 7c have the same configuration as the music distribution site 7a.

(Operation) First, a member registration procedure for a user to receive a music selection service will be described.

The user operates the web server 64 of the music selection site 6 using the mobile phone 1 having an Internet connection function.
Connect to and open the member registration page linked to the homepage. On the member registration page, a form (screen) including an input box for inputting the user's name, current address, gender, age, telephone number, e-mail address, and credit card number is displayed. Then, the user operates the confirmation button and the like to sequentially input and transmit the input to the web server 64. The WEB server 64 determines whether or not such information has been input (whether or not the information is not null). If the determination is negative, the web server 64 notifies the user of the missing input item and prompts the user to input the information. Is transmitted to the mobile phone 1 via the packet center 3 to display a form including an input box for allowing the user to input the form on the display of the mobile phone 1 to prompt the user to input. The WEB server 64 waits until a password is input. When the password is input, the WEB server 64 captures the password, issues a unique user ID, and then issues a music selection server 65 of the music selection site 6 or a music distribution site. Then, an application program described in JAVA for accessing the music distribution server 73 (hereinafter, referred to as a JAVA program) is transmitted to the mobile phone 1, and the member registration procedure is completed. As a result, the JAVA program is stored in the mobile phone 1, and the user calls the JAVA program in which the URL information of the music selection server 65 is described, and inputs a user ID and a password to enter the member site of the music selection server 65. It is possible to enter. After transmitting the JAVA program, the WEB server 64 transfers the data such as the user's name, the user ID and the password to the database server 66 for storage.

Next, the operation of the music selection site will be described together with the operation of the user and the music distribution site with reference to a flowchart.

The user operates the numeric keypad or the like to invoke the JAVA program stored in the mobile phone 1 and activates it. The display of the mobile phone 1 indicates that the pulse wave measuring device 2 should measure (detect) a pulse wave. The user
After turning on a power switch (not shown) of the pulse wave measuring device 2 to attach the pulse wave sensor 21 to, for example, a forefinger, the enter button is pressed. Thus, the measurement of the pulse wave in the initial state of the user is started (see t1 in FIG. 11). That is, by allowing the CPU of the mobile phone 1 to establish a wireless link between the BTM in the mobile phone 1 and the BTM 24 in the pulse wave meter 2 according to the JAVA program, the pulse wave meter 2
The reception of time-series pulse wave data transmitted at a sampling rate of 000 Hz is started, and the received pulse wave data is stored in the SD memory in the mobile phone. CPU of mobile phone 1
Determines whether or not 40 seconds have elapsed. If the determination is negative, the reception is continued. If the determination is affirmative, the reception is terminated (see t2 in FIG. 11). The wireless link with the BTM 24 in the device 2 is terminated. Thus, the number of data stored in the SD memory becomes 40,000.

Next, the user accesses the music selection server 65 by the JAVA program as described above. The music selection server 65 accessed from the mobile phone 1
As shown in FIG. 6, a music selection routine for selecting the most suitable music for achieving the psychological state desired by the user is executed.

In the music selection routine, first, step 1
At 02, an authentication process is executed. In this authentication process,
A form in which an input box for allowing a user to input a user ID and a password is displayed on the display of the mobile phone 1 and waits for a response from the mobile phone 1. , User I
D. The password is fetched, the database server 66 determines whether the corresponding data exists, and responds.
If the database server 66 responds that the corresponding data is present, it is authenticated that the user is a member, so the process proceeds to step 104. If the database server 66 responds that the corresponding data does not exist, a document prompting input is displayed again. And wait until a response is received.

In the next step 104, a large category of the mental state desired by the user is displayed on the display of the mobile phone 1, and in step 106, the process waits until the user selects one of the large categories. The major categories are stress relief, refreshment, healing of sorrow, rejuvenation,
It is classified into six types: anxiety relief and a good night's sleep. Step 10
If an affirmative decision is made in step 6, in the next step 108, a default value for specifying the selected large category is taken in, and the user inputs taste data such as a favorite music genre and singer's gender (hereinafter, referred to as a mode). In order to inquire whether the mode is an additional mode) or a mode in which no preference data is input (hereinafter referred to as a non-additional mode), a screen prompting the user to select a mode is displayed on the display of the mobile phone 1. .

Next, at step 110, the process waits until there is a selection by the user, and if there is a selection, it is determined at step 112 whether or not the mode is the addition mode.
At 14, an additional mode processing subroutine is executed.

As shown in FIG. 7, in the additional mode processing subroutine, first, in step 122, a form describing the detailed classification belonging to the large classification is transmitted to the mobile phone 1 and displayed on the display. It waits until there is a selection response from the mobile phone 1. When the major category is "stress release", for example, the detailed category is "angry by the boss of the company", "noisy wife", "family relationship does not go well", "friends go wrong", " You are asked to select one of the following seven: "I am bullying", "I do not work well" or "I cannot study for the entrance exam".

When there is a selection response from the mobile phone 1 in step 124, the detailed classification selected in the next step 126 is fetched, and the preference data fetching process is executed. In this preference data capturing process, the genre, singer's gender, singer, lyricist / composer, and music release date, which are preference data representing the user's favorite music data. Display a form prompting the user to enter the mobile phone 1
Wait for a response from. "Genre" can be selected from pops, rock, jazz, classical music, and enka. "Singer's gender" can be selected from men and women. The "singer" and the "lyricist / composer" can be searched by the Japanese syllabary by supporting the music selection server 65 (for example, when the first or second character of the singer is input, the character is changed to the corresponding character). All the corresponding singers are displayed on the display in the order of the Japanese syllabary). "Music release age"
Then, designation such as the 1970s is possible. The user does not have to input the preference data that he does not want to input, or that cannot be determined or is unknown. When there is a response from the mobile phone 1, the preference data is fetched.

Next, at step 128, step 126
The detailed categorization and preference data fetched in step (1) are sent to the database server 66 to search for a suitable song and make it respond. As shown in FIG. 12, the database server 66 (and the analysis server 67) includes major classification, detailed classification, stress (A), sadness (B), joy (C), and relaxation (D).
Is stored. A method for creating this state table will be described later. For example, when the detailed classification selected by the user is “angry by the boss of the company”, the database server 66 refers to the state table and decreases A (↓↓) and B decreases (↓↓).
↓), a combination in which C increases (↑↑) and a small increase in D (と す る) are search targets. Note that the same detailed classification also exists in different combinations such as “My wife's bitches are noisy”. In addition, the database server 66 stores, for each song, the genre, singer's gender, singer, lyricist / composer, and music presentation, which are the above-mentioned preference data as attribute data of the song. In addition to the “age”, additional music data of “applicable season (whether it is suitable for spring, summer, autumn and winter or plural or more)”, “release date”, and “whether or not campaign music” is stored. Accordingly, the database server 66 first searches for the corresponding song from the preference data, determines whether or not the attribute data of the corresponding song matches the search target in the state table. If the answer is affirmative, the matching song name and its URL are stored in the RAM, and a similar search is performed for all the songs stored in the database server 66. When the search for all the songs is completed, the compatible song name stored in the RAM and its URL are sent to the music selection server 65. Music selection server 65
Counts the number of compatible songs.

In the next step 130, the counter q is set to 1
Is determined. If the determination is affirmative, step 13
6; if negative, the next step 132 is to judge whether the number of compatible songs is 5 (tunes) or more; if affirmative, proceed to step 138; At 134, the preference data range expansion acquisition process is executed, and the process returns to step 128. This preference data range expansion capture process displays the current number of compatible songs on the display of the mobile phone 1 in order to correct that the number of compatible songs is less than five in the preference data capture range captured in step 126. Then, as in step 126, an input for expanding the range of the preference data is obtained again. In order to expand the range of the preference data, for example, it is possible to respond by deleting the input of "genre" which is a part of the preference data, or by inputting two to three or more "singers". .

In step 134, the above-mentioned counter q is changed from "0" to "1" in order to give the user only one chance to expand the range of the preference data. Therefore, when an affirmative determination is made in step 130, the opportunity to expand the range of the preference data has already been given. In order to shorten the communication time with the mobile phone 1, in step 136, the database server 66 Add music for the missing songs. That is, the music selection server 65 sends the number of missing songs to the database server 66, and among the songs corresponding to the campaign song from the song additional data, the above-described search target and “matching season” match the current season (or Select the music that matches the current season), and randomly select the appropriate song titles and U
By responding to the RL, the responding tune name and the URL that responded are additionally stored in the RAM, and the process proceeds to step 140. In the database server 66, the data of the “campaign music” is updated weekly by a program that sets six months after the “release date” as “good or bad for the campaign song” and no more than six months. .

On the other hand, in step 138, since there are five or more compatible songs, five of them are randomly selected from among them to be adapted songs. Next, in step 140, the titles of the five songs and their URLs are transmitted to the mobile phone 1, and the process waits until measurement data is received from the mobile phone 1 in the next step 142.

Upon receiving the song titles and URLs of the five songs, the mobile phone 1 creates a form including a trial listening button for starting the trial listening of the selected song in which the received song titles are described according to the JAVA program. Display on the display. The user operates the cursor direction button and the confirm button to click the preview button (see t3 in FIG. 11).
As a result, the mobile phone 1 sequentially accesses the transmitted URLs of the five songs. For example, the case where the URL of the first song corresponds to the music distribution site 7a will be described in detail. The music distribution server 73 stores a host music server name and address correspondence table, and according to the received URL, for example, the music server Music data for trial listening (music data obtained by extracting a part of full music data) stored at a predetermined address of the music server 74 is transmitted from the music server 74 to the mobile phone 1 in a streaming format, for example, in a Real format file (FIG. 11 t4). Therefore, the music distribution server 73 in this case functions as a so-called DNS server. Note that streaming is a technology for delivering multimedia files online. In the case of download format delivery, the entire multimedia file is temporarily stored in memory and then played back. The received multimedia files are sequentially reproduced without downloading. The playback time of the compatible tune transmitted in the streaming format is set between 30 seconds and 100 seconds.

The mobile phone 1 starts receiving Real-format files according to the JAVA program, and when a certain amount of data is stored in the RAM of the mobile phone 1 functioning as a buffer, the music received by packet communication is sequentially transmitted. It reproduces and outputs it to the speaker (or earphone) of the mobile phone 1. At the start of music playback (see t5 in FIG. 11),
The JAVA program allows the establishment of a wireless link between the BTM in the mobile phone 1 and the BTM 24 in the pulse wave meter 2, and this wireless link is established after the end of the reproduction of the transmitted music (see t6 in FIG. 11). This is performed until 2 seconds later (t8 in FIG. 11).
reference). As described above, the playback time of the transmitted music is 3
The time is set to 0 to 100 seconds, and the pulse wave measuring device 2 measures pulse wave data for 2 seconds thereafter (total 32 seconds to 102 seconds).
Second), the sampling rate of the A / D converter 23 is 1, 0
00 Hz, the U
The IM module has 32,000 to 10 songs per song.
2,000 pulse wave data are stored.

The mobile phone 1 accesses the URL of the second music at time t7 (time t6 + 1.5 seconds) in FIG. 11 according to the JAVA program. At this time, U
Since the RL is not always the same as the URL of the first song, the user accesses one of the music distribution sites 7.
As in the case of the first song described above, at time t9, the music server 74 transmits music data for trial listening to the mobile phone 1 in a streaming format. At time t10, the user listens to the second song and listens to the pulse wave meter 2 for the user. Measurement of the pulse wave is started. Similarly, the listening to the fifth song and the measurement of the pulse wave are performed, and the listening to the fifth song by the user ends at time t11.
At time t12, the measurement of the pulse wave of the fifth music is completed. At this point, the UIM attached to the mobile phone 1 stores the measurement data in the initial state and the measurement data of the first to fifth songs. From time t13, these measurement data are transmitted to the music selection site 6.

Therefore, in step 142 of FIG. 7 described above, the process waits until the measurement data is received. When the measurement data is received, the received measurement data is transferred to the analysis server 67 in the next step 144. ID to specify
The number and the user ID are transmitted, and in step 146, the listening music data is transmitted to the mobile phone 1 in a streaming format. The music data for trial listening here is music data for trial listening of a song randomly selected from campaign / promotion songs specially contracted with a record company. In the next step 148, it is determined whether or not there is a response (selection of the optimum music) from the analysis server 67. If a negative determination is made, the process returns to step 146; Accordingly, the user waits for a response (selection of the optimum music) from the music selection site 6 while listening to the campaign / promotion music played on the mobile phone 1 during this time.

During this time, as shown in FIG. 8, the analysis server 67 analyzes the measurement data and selects an optimum music having a psychological state desired by the user from the suitable music according to the current mental state of the user. Execute the analysis / selection routine.

In the analysis / selection routine, first, in step 202, the measurement data transferred from the music selection server 65 and the ID numbers and user IDs of five pieces of music are taken in (corresponding to step 144 in FIG. 7), n
Is set to 6. The processing number counter n corresponds to the number of blocks of the measurement data, and is obtained by adding the measurement data blocks of the first to fifth music pieces to the above-described measurement data block in the initial state. In the next step 204, it is determined whether or not the number-of-processes counter n is 0. If the determination is negative, the autocorrelation R (L) is calculated in the next step 206.

Here, chaos used in this embodiment will be described.
Will be explained. It was disclosed in Japanese Patent Publication No. 6-9546
As is clear from technology, pulse waves can be chaotic
Are known. Therefore, chaos is obtained from the time series data of pulse wave.
It is possible to obtain features of attractors. Chaotic
Definitions vary by researcher and various definitions have been proposed.
However, in the present embodiment, according to the definition of Ott (Ott),
The solution is not periodic, the autocorrelation function is the lag time
Sensitive to the initial value, which converges to 0 with increasing
Chaos if it meets the requirement of
(Strangeattractors and chaotic motions of
dynamical systems, Rev. Modern Phys., 58 (4/1), pp.6
55-671 (1981)). In the following steps we will define the otto
Therefore, the pulse wave is chaotic and the
Consideration of period (for example, Fast Fourier Transform)
 Transform) is not necessary.
Autocorrelation analysis (calculation of R (L) described later)
Noff exponential analysis (λ ₁Calculation) to perform chaos
The feature amount of the attractor is obtained.

In step 206, an autocorrelation R (L) represented by the following equation 1 is obtained for each delay time L.

[0043]

(Equation 1)

FIGS. 13A and 13B show the autocorrelation R (L) every 1 ms from L = 1 ms to 100 ms by taking the delay time L on the horizontal axis and the autocorrelation R (L) on the vertical axis. 5 schematically shows the relationship between the delay time L and the autocorrelation R (L) in the case of calculation. Autocorrelation R in the case of chaos
(L) may take a negative value as shown in FIG. 13 (A) and converge to 0 as shown in FIG. 13 (B), or a positive value and converge to 0 as shown in FIG. 13 (B). is there.

In the next step 208, the autocorrelation R
It is determined whether (L) is smaller than 0. If the determination is affirmative, the autocorrelation R
The delay time L _{0 at} which (L) becomes ₀ is determined (see FIG. 13A). The embedded delay time τ is set to the delay time L ₀ , and the process proceeds to the next step 214. On the other hand, in the case of negative determination, initially as the autocorrelation R (L) is determined a delay time L ₁ which takes a minimum value (Fig. 13 (B) refer) embedding delay time tau = time delay L ₁ in step 212, Next step 214
Proceed to.

In the next step 214, the Takens (Tak
ens), the correlation integral C (r) is obtained. Takensu of embedding theorem, when a guaranteed theorem that it is possible to combine the presence of the sampling results and attractor series data, that attractor of the dynamical system is rebuilt is smoothly converted into R ^{2m + 1} by the mapping V means. If it is chaos, it generally has a fractal structure, and the characteristics of a chaos attractor can be extracted as a fractal dimension. The fractal dimension is an extension of the normal dimension concept to non-integer domains, and includes Hausdorff dimension, correlation dimension (GP
Law) has been proposed. In the present embodiment, the correlation integral C (r) defined by the following equation 2 from two dimensions to m dimensions is calculated using the most commonly used correlation dimension method. In Equation 2, the step function H of Heaviside is
(Y) is a function that takes 1 when y> = 0 and takes 0 when y <0, and the absolute value of x _i −x _j represents the Euclidean distance of the vector x.

[0047]

(Equation 2)

The logarithm logC (r) of the correlation integral C (r) of each embedding dimension calculated in step 214 is plotted on the vertical axis.
FIG. 14 schematically shows the logarithm logr of r with the horizontal axis.
It becomes as shown in. This correlation integral C (r) is C (r) ∝
^D when scaled like r ^D is called the correlation index. The correlation index D is represented as a slope in FIG.
Therefore, in the next step 216, a correlation index D for each embedding dimension d is calculated. As shown in FIG. 15, in the case of chaos, the correlation index D plotted for each embedding dimension d saturates (converges) at a certain correlation dimension Dc. In step 216, the embedding dimension d and the correlation dimension Dc at this time are acquired. If the determination method in the present embodiment is explained based on the example shown in FIG. 15, 4 at which the correlation index D starts to be saturated is set as the embedding dimension d, and the correlation index D at that time is set as the correlation dimension Dc.

In the next step 218, the Lyapunov exponent λ ₁ is calculated for each embedding dimension. The Lyapunov exponent is a time n → ∞ of two trajectories starting from two close points in a dynamical system represented by x _{n + 1} = f (x _n ).
, And is defined by the following equation (3).

[0050]

(Equation 3)

FIG. 16 schematically shows a Lyapunov spectrum of Lyapunov exponents λ ₁ to λ ₄ from d = 1 to d = 4 when the embedding dimension d = 4 is determined. In step 218, it calculates only Lyapunov exponent lambda ₁ that most affect later to state vector takes the maximum value among these Lyapunov exponent. Next step 2
At 20, the Lyapunov exponent λ ₁ and the fractal dimension (correlation dimension) Dc are stored in the RAM, and at step 222, 1 is decremented from the processing number counter n, and the process returns to step 204.

On the other hand, if an affirmative decision is made in step 204, in step 224, the user ID fetched in step 202 is sent to the database server 66 to search and respond to the user's age a and gender s. The transmitted user's age a and gender s are stored in the RAM.

In the next step 226, the analysis server 67
Call the neural network program stored in. The neural network of this embodiment is
As shown in FIG. 17, it has a three-layer structure including an input layer, an intermediate layer, and an output layer, and inputs a state vector (λ ₁ , Dc, a, s), which is a characteristic amount of the chaos attractor, to the input layer. By doing so, sensibility vectors (A, B, C, D) representing stress (A), sadness (B), joy (C), and relaxation (D) are output from the output layer. In addition,
In the present embodiment, the state vector (λ ₁ , Dc, a, s)
Lyapunov exponent lambda ₁ constituting the _fractal dimension (correlation dimension) Dc, age a and sex s is normalized data are used between 0-1.

Here, the relationship between the neural network, the state vector (λ ₁ , Dc, a, s) and the sensitivity vector (A, B, C, D) will be outlined. A graph in which each coordinate axis is essentially associated with one dynamic variable is called a state space. One point in the state space represents the state of the system at a given time. Chaotic systems have complex trajectories in state space, whereas linear systems have simple trajectories such as closed surfaces. The trajectory of a chaotic system passes only through a specific region of the state space and does not pass through other regions. The trajectory describes a chaotic attractor. This chaos attractor can be reconfigured by embedding time-series data having chaos in a multidimensional state space. For embedding, n state variables may be restored from one state variable by using the embedding delay time τ according to the Turns method. If the embedding can be performed, the reconstructed chaotic attractor is a modification of the original attractor, and the Lyapunov exponent λ ₁ , the fractal dimension Dc, and the like are topologically preserved. In order to reconstruct the chaos attractor from the univariate time-series data, embedding by conversion into a delay time coordinate system as in the following Expression 4 may be performed.

[0055]

(Equation 4)

Where X (t) is univariate observation time series data, d is an embedding dimension, τ is an embedding delay time, and V
(T) represents the transformed reconstructed state space data. Chaos has stability such that even if orbital instability is caused by a small disturbance, the geometric structure of attractors that exhibits steady-state behavior in the state space does not change ("Time-series analysis system based on deterministic chaos theory"). "Instrumentation August Vol. 40, No. 8 (1997)). Therefore, in the neural network according to the present embodiment, as described above, the state vector (λ ₁ , Dc, a, s) is input to the input layer, and the stable sensitivity vectors (A, B, C,
D).

Subsequently, learning by the neural network in this embodiment, that is, the state vector (λ ₁ , D
A description will be given of a method of correlating the sensitivity vector (A, B, C, D) with the sensitivity vector (A, B, C, D) and a method of creating the state table shown in FIG. As shown in FIG. 11, the pulse wave measuring device 2 measures pulse wave data of a certain subject in an initial state (before listening to music) and pulse wave data when listening to the first to fifth music. I do. At this time, stress (A), sadness (B), joy (C),
A questionnaire is taken for relaxation (D), and the results of the questionnaire are normalized between 0 and 1 for stress (A), sadness (B), joy (C), and relaxation (D). According to the same procedure as in Steps 204 to 224 shown in FIG. 8, the state vector (λ ₁ , Dc, a, s) of the subject is, for example, (0.78,
0.32, 0.41, 1.00). This state vector (λ ₁ , Dc, a, s) is input to the input layer of the neural network shown in FIG. 17, and the correct sensitivity vector (A, B, C, D) is obtained (0.78, 0.32). ,
0.41, 1.00) is trained by the learning rule of back propagation. By learning about 100 subjects, even if the state vector of an unknown subject is input, the sensitivity vector of the subject can be output. The neural network of the present embodiment is completed by learning 100 subjects, and the neural network program is obtained by transplanting the neural network.

For a subject with a high stress level, for example, 5 songs
Listen to the music. At this time,
A questionnaire survey was conducted to determine whether or not divergence occurred. That strain
Vector when listening to a song where
Vector (before listening to music)
It is called sensitivity vector. ) And calculate the difference between
(A), sadness (B), joy (C), relax (D)
Check how things are changing. That is,
The sensitivity vector (A, B, C, D) of the subject is, for example,
Initial sensitivity vector (A₀, B₀, C₀, D₀) = (0.
70, 0.45, 0.11, 0.08) Listen to the first song
Sensitivity vector (A₁, B₁, C₁, D₁) =
(0.55, 0.19, 0.21, 0.08), 2nd song
Vector (A) when listening to₂, B₂, C₂, D
₂) = (0.88,0.45,0.11,0.08),
Sensitivity vector (A₃, B₃, C
₃, D ₃) = (0.70, 0.75, 0.11, 0.0
8) The sensitivity vector (A₄, B
₄, C₄, D₄) = (0.14, 0.19, 0.41,
0.28) Sensitivity vector (A
₅, B₅, C₅, D₅) = (0.14, 0.19, 0.
35, 0.08), the initial sensitivity vector (A
₀, B₀, C₀, D₀) And the first song's sensitivity vector
(A₁, B₁, C₁, D ₁) And the difference (A₀-A₁, B₀
-B₁, C₀-C₁, D₀-D₁) = (0.15,0.
26, -0.10, 0.00). Similarly,
Initial sensitivity vector (A₀, B₀, C₀, D₀) And the second song
Sensitivity vector (A₂, B₂, C₂, D₂), The first
Period sensitivity vector (A₀, B₀, C₀, D₀) And the third song
Sensitivity vector (A₃, B₃, C₃, D₃) Difference, initial
Sensitivity vector (A₀, B₀, C₀, D₀) And the feeling of the fourth song
Sex vector (A₄, B₄, C₄, D₄), Initial feeling
Sex vector (A₀, B₀, C₀, D₀) And the sensitivity of the fifth song
Vector (A₅, B₅, C₅, D₅Calculate the difference with
You. Then, by performing the same procedure on multiple subjects,
General stress to release tress (A), sad
Combination of change of only (B), joy (C), relax (D)
Knowledge about alignment can be obtained. That is, FIG.
One line shown in FIG. 2 is obtained.

The criterion for selecting which of the first to fifth songs fits the psychological state desired by the user is as follows.
It is whether or not the most appropriate change has been made from the initial sensitivity vector. The change in the initial sensitivity vector does not mean that the component difference (for example, A ₀ −A ₁ ) simply decreases (or increases). "Stress (A) decreases (↓
↓), sadness (B) increases (↑↑), joy (C) decreases slightly (↓), and relaxation (D) increases (↑↑). It should be noted that the relation with the arrows and the like in the state table shown in FIG.
For example, when the difference between the components exceeds 0.15,
“↑” when 0.05 to 0.15, 0.05 (less than)
"-" For ~ -0.05 (-0.05 is not included), "↓" for -0.05 to -0.15,-
A response is obtained by setting "↓↓" when the value is smaller than 0.15.

The record company that operates the music distribution site 7 has newly established a music selection site 6 for about three subjects.
The combination of sensibility vectors for new songs and the like stored in the database server 66 is calculated, and along with this combination, "genre", "sex of singer", "singer", "lyricist / composer", "music release date", The data relating to the “suitable season” and the “release date” are transmitted to the music selection site 6 via the Internet 5 in a predetermined format. Music selection server 6
Reference numeral 5 designates an ID number for each song, and also assigns attribute data of a large classification and a detailed classification to the new song or the like from the state table shown in FIG.
6 includes “genre”, “singer's gender”, “singer”, “lyricist / composer”, “music release date”, “adapted season”,
Store with "Release date". It should be noted that, at this time, "whether or not to use a campaign song" is updated by a separate program every week, so that it is not necessary to store it in the database server 66.

In step 228 of FIG. 8, the state vectors (λ ₁ , Dc, a, s) for five songs are input to the neural network program, and in the next step 230, the sensitivity vectors (A, B, C, D) are calculated. Next, in step 232, as described above, the difference between the initial kansei vector and each kansei vector is calculated, and a combination of kansei vectors (A, B, C, D) for five music pieces is obtained.

In the next step 234, first, since the combination of the sensibility vectors from the music distribution site 7 is performed by about three subjects, it may be necessary to correct the combination. Among the combinations of the sensibility vectors (A, B, C, D), those that do not match the combinations of the state tables corresponding to the detailed classification are excluded. Such correction may be performed, for example, by periodically notifying the music distribution site 7 from the analysis server 67. Next, among the combinations of “stress (A), sadness (B), joy (C), and relaxation (D)” in the state table, the emotion corresponding to “↑↑” and “↓↓” is the psychological state desired by the user. Characteristic, so of the differences between the sensitivity vectors that match the combination of the initial sensitivity vector and the state table, the sum of the squares of the component values corresponding to “↑↑” and “↓↓” is calculated. The song corresponding to the one with the large numerical value is selected as the optimal song. For example, the combination of “stress (A), sadness (B), joy (C), and relaxation (D)” in the state table corresponding to the psychological state desired by the user is “↑↑,
↓,-, ↓↓ ", and when the first, third, and fifth tunes match this combination, the difference あ る (A ₀ −A ₁ ) between the component values of“ ↑↑ ”,“ ↓↓ ” ² + (D ₀ −
D ₁ ) ² }, {(A ₀ −A ₃ ) ² + (D ₀ −
D ₃ ) ² }, {(A ₀ −A ₅ ) ² + (D ₀ −D ₅ ) ² }
Is calculated, and the tune corresponding to the largest numerical value among these calculation results is selected as the optimum tune. In step 236, the ID number of the optimum music is transmitted to the music selection server 65, and the analysis / selection routine ends.

When the music selection server 65 receives a response from the analysis server 67, in step 150 in FIG. 7, the music selection server 65 fetches the ID number of the optimum music, describes the optimum music name, and executes a program for accessing the URL of the optimum music. The form in which the allocated purchase button is arranged is transmitted to the mobile phone 1, and the additional mode processing subroutine is terminated, and the process returns to step 114 in FIG. 6 to terminate the music selection routine. Thereby, the user can know the optimal music for obtaining the desired mental state by looking at the display on the display of the mobile phone 1. By pressing the purchase button, it is possible to access the URL of the optimal song, download the optimal song to improve the sound quality, and store it in the UIM. The music distribution server 73 transmits a form describing notes on purchase to the mobile phone 1. In this form, an agreement button is arranged. When the agreement button is pressed by the confirmation button, the music distribution server 73 causes the accounting server 76 to store the accounting information. The billing server 76 calculates a monthly billing amount for each member and notifies the music selection server 6 of the billing amount. By reproducing and listening to the optimum music downloaded to the UIM, the user can obtain a desired psychological state.

On the other hand, when a negative determination is made in step 112 of FIG. 6, in step 116, the non-additional mode processing subroutine shown in FIG. 9 is executed, and the music selection routine ends. This non-addition mode is based on the preference data in advance, in order to prevent seemingly unacceptable music selection in the deep psychology even though the user thinks that the user likes a certain music (the body does not respond). This is a mode in which is not limited. 9 and 10, the same steps as those in FIGS. 7 and 8 are denoted by the same reference numerals, and the description thereof will be omitted. Only different points or steps will be described.

In step 162 of FIG. 9, among the songs belonging to the large category taken in step 108 of FIG. 6, ten songs having different tendencies (tunes having different detailed categories) are selected. For example, when there are seven detailed classifications belonging to the large classification, at least one song is randomly selected from each detailed classification to make a total of 10 songs. Such a selection is made, for example, by selecting, from among the campaign songs of each detailed category belonging to the large category, songs that match the current season one by one, randomly selecting the detailed category, and similarly selecting the remaining three songs. By doing so, it is possible to select 10 songs. Therefore, in the non-addition mode, the music selection server 65 transmits the URLs of the ten songs to the mobile phone 1 (step 140), and the user listens to the ten songs.

The analysis server 67 determines in step 20 in FIG.
In step 2, the processing number counter n is set to 11 and step 2
Steps after step 04 are executed. Step 226
The neural network program called by
A neural network that converts a state vector into a sensitivity vector and is learned for the non-addition mode is transplanted (the creation method and the like are the same as those for the addition mode described above). Further, in step 302, since the user has not selected the detailed classification in the non-addition mode, among the combinations of the sensitivity vectors (A, B, C, D) for the ten songs obtained in step 232, Exclude ones that do not match the combination of any of the state tables of the detailed classification (plural) belonging to the classification. Next, among the differences between the initial emotion vector and the emotion vector corresponding to the combination of the state table to which the music corresponds, “↑↑”, “↑”,
The sum of the squares of the component values corresponding to “↓” and “↓↓” is calculated, and the song corresponding to the largest value is selected as the optimal song. In the next step 304, the I
The detailed classification is sent to the music selection server 65 in addition to the D number, and the analysis / selection routine ends. Music selection server 6
5 receives a response from the analysis server 67 and returns to step 150
In, capture the ID number and detailed classification of the optimal song,
A form in which the name of the optimum music and the detailed classification are described, and a purchase button to which a program for accessing the URL of the optimum music is allocated is transmitted to the mobile phone 1, and the non-additional mode processing subroutine is terminated, and the steps in FIG. Return to 116.

According to the present embodiment, the state vector is calculated from the pulse wave data (steps 202 to 224), the sensitivity vector is obtained by the neural network (steps 226 to 230), and the initial sensitivity vector and each sensitivity vector are obtained. Is calculated to select the optimum music (steps 232 and 234), it is possible to select a music suitable for the mental and physical condition desired by the user. Also,
The pulse wave measuring device 2 has a simple configuration, and the mobile phone 1 can access the music selection site 6 and the music distribution site 7, and the operation for selecting music is performed on the site side, so that appropriate music can be easily selected. It becomes possible.

In the present embodiment, an example in which the mobile phone 1 accesses the music distribution site 7 has been described. However, as shown in FIG. 1, a personal computer 9 connects to the provider 8 via a public line, and The music selection service may be provided.

In this embodiment, the music selection server 6
5 shows an example in which the music data is transmitted to the mobile phone 1 in step 146, but an image advertisement that is considered optimal based on the user ID may be transmitted to the mobile phone 1 at the same time.
In addition, the music selection server 65 provides the user with information on the environment in which the mobile phone 1 listens to the optimal music, such as room temperature adjustment, lighting intensity, massage chair vibration pattern, whirlpool flow rate control, and the like. May be led to a more comfortable environment. Further, when the user selects the detailed classification in the additional mode, the music selection server 65 may send an advice that it is better to sing along with the trial listening according to the detailed classification.

Further, in the present embodiment, the case where the analysis / selection by the analysis server is performed online has been exemplified. However, the communication is temporarily terminated before the analysis / selection routine is started, and the mobile phone is re-executed after a predetermined time. It is also possible to access from one side and receive the notification of the optimum music.

In this embodiment, the music distribution server 7
3 shows an example in which only music data is transmitted, for example, when transmitting music data for trial listening, a landscape image considered to be most suitable for trial listening to the music may be attached at the same time. In this way, it is possible to give the user a better image of the preview music.

Further, the music selection site 6 obtains information about the user's physical characteristics such as height and weight, whether work stress is accumulated, and the like through a questionnaire after the music selection is completed. It is possible to obtain statistics on music that the member feels comfortable in consideration of the social environment, and to provide a more accurate music selection service by increasing the number of statistics.

[0073]

As described above, according to the present invention,
Among the increasing / decreasing trends for the calculated difference, a song having an increasing / decreasing tendency that matches the increasing / decreasing tendency of the sensitivity vector set in advance corresponding to the song is selected as a song for which the subject can obtain the desired emotional state. Thus, it is possible to obtain an effect that a music piece suitable for the mental and physical condition desired by the subject can be selected according to the state before listening to the music of the subject.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of each site of an embodiment in which a music selection service to which the present invention can be applied is executable.

FIG. 2 is a block diagram illustrating a schematic configuration of a pulse wave measuring device according to the embodiment.

3A and 3B show a pulse wave sensor of the pulse wave measuring device, wherein FIG. 3A is a sectional view and FIG. 3B is a circuit diagram.

FIG. 4 is a configuration diagram showing an outline of a music selection site.

FIG. 5 is a configuration diagram schematically showing a music distribution site.

FIG. 6 is a flowchart of a music selection routine executed by a music selection server of a music selection site.

FIG. 7 is a flowchart of an additional mode processing subroutine of step 114 of the music selection routine.

FIG. 8 is a flowchart of an analysis / selection routine executed by the analysis server when the music selection server is executing the additional mode processing subroutine.

FIG. 9 is a flowchart of a non-addition mode processing subroutine of step 116 of the music selection routine.

FIG. 10 shows an analysis and analysis performed by the analysis server when the music selection server is executing the non-additional mode processing subroutine.
It is a flowchart of a selection routine.

FIG. 11 is a timing chart showing the operation timing of the mobile phone.

FIG. 12 is a conceptual diagram of a state table indicating a tendency to increase or decrease by arrows of large classification, detailed classification, sensitivity vector elements, and arrows.

FIG. 13 is an explanatory diagram schematically showing the autocorrelation when the delay time is plotted on the horizontal axis and the autocorrelation is plotted on the vertical axis. FIG. 13A shows a case where the autocorrelation sometimes takes a negative value. (B) shows a case where the autocorrelation takes a positive value.

FIG. 14 is a correlation diagram of correlation integration when Logr is plotted on the horizontal axis and LogC (r) is plotted on the vertical axis.

FIG. 15 is a convergence diagram of the correlation dimension when the embedding dimension is plotted on the horizontal axis and the correlation index is plotted on the vertical axis.

FIG. 16 is a Lyapunov spectrum in which Lyapunov exponents when the embedding dimension is 4 are arranged in descending order.

FIG. 17 is a structural explanatory view showing a schematic structure of a neural network.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Cellular phone 2 Pulse wave meter 3 Packet center 5 Internet 6 Music selection site 7a, 7b, 7c Music distribution site 65 Music selection server 66 Database server 67 Analysis server 73 Music distribution server 74, 75 Music server

Claims (6)

[Claims] 1. A music selection support method for supporting selection of a desired music from a large number of music, wherein the pulse wave time-series data before listening to a music of a subject and a plurality of music data when a plurality of music are being heard. Calculate the state vector representing the characteristic amount of the chaos attractor from the pulse wave time series data of the above, input the state vector to the input layer of the neural network, and listen to the pulse wave time series data before listening to the music and a plurality of music pieces Is obtained from the output layer for each of the plurality of pulse wave time-series data when listening to the sensibility vector of the pulse wave time-series data and the tune before listening to the tune. Calculate the increase / decrease tendency for the difference with the sensitivity vector of the pulse wave time-series data when it is, and among the increase / decrease trends for the calculated difference, the tendency is set in advance corresponding to the music. A music selection support method including the step of selecting a music piece having an increasing / decreasing tendency coinciding with the increasing / decreasing tendency of the sentiment vector as a music piece having a desired emotional state desired by the subject. 2. The music selection support method according to claim 1, wherein the elements of the state vector include at least a Lyapunov exponent and a correlation dimension. 3. The music selection support method according to claim 1, wherein the sensitivity vector includes at least a sensitivity element of stress, sadness, joy, and relaxation. 4. Receiving pulse wave time-series data before listening to the subject's music from a moving object and a plurality of pulse wave time-series data when listening to a plurality of music, and Any one of claims 1 to 3
Music selection support method described in section. 5. A program that allows a computer to execute the music selection support method according to any one of claims 1 to 4. 6. A computer-readable recording medium in which the program according to claim 5 is described.