A music scene recognition system based on artificial intelligence

Abstract

The present invention relates to the field of data analysis system computing, and specifically to a music scene recognition system based on artificial intelligence, which at least includes: a monitoring terminal for monitoring human body data, a playback terminal for playing music, and also includes: a state recognition unit for identifying the user's physical state, a habit analysis unit for analyzing the user's behavior, a scene determination unit for determining the current music scene, and a song list construction unit for constructing a user's personalized recommended song list. The present invention determines whether the user is currently in a habitual behavior scene by analyzing and comparing the time interval of the user's opening moment, thereby inferring the user's music scene, and, when in a fuzzy interval, increases the state verification, thereby improving the accuracy of scene analysis when it is uncertain whether the user is in a habitual behavior scene.

Description

Technical Field

[0001] The present invention relates to the field of data analysis system computing, and in particular to a music scene recognition system based on artificial intelligence. Background Art

[0002] The accompanying nature of music itself naturally diversifies its application scenarios, and the application of technologies such as artificial intelligence has made it possible to further extend the application scenarios of music. Data shows that in-car music, smart home and smart speakers are the top three devices used by music users, accounting for 38.4%, 37.0% and 36.8% respectively. The improvement and integration of new technologies and the continuous extension of music application scenarios have enhanced the "strong accompaniment" attribute of music, with travel and home being the main scenarios.

[0003] Although major music software platforms have set up different scenario playlists, these scenario playlists often require users to actively choose to play or are directly promoted by the music software platform to all users' interactive interfaces. This makes it difficult for the music software platform's scenario playlists to adapt to users' actual needs, and even more difficult to recommend corresponding playlists to users in appropriate music scenarios, affecting users' listening experience in different scenarios. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, the present invention provides a music scene recognition system based on artificial intelligence, which can effectively solve the problem of difficulty in music scene recognition when music software in the prior art recommends music to users.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0006] The present invention provides a music scene recognition system based on artificial intelligence, which at least includes: a monitoring terminal for monitoring human body data, a playing terminal for playing music, and also includes:

[0007] A state recognition unit for identifying the user's physical state, obtaining and analyzing the user's body data, and marking the user as being in a moving state, a resting state, or a driving state;

[0008] The habit analysis unit for analyzing user behavior obtains the user's historical status data and defines three behavior intervals, namely driving interval, exercise interval and rest interval, corresponding to driving, exercise and rest. The overlap of user behaviors is clarified according to the distribution of user behavior intervals, thereby constructing standard intervals and fuzzy intervals for various behavior distributions;

[0009] The scene determination unit is used to determine the current music scene. It records the moment when the user turns on the playback terminal as the opening time, compares the opening time with each standard interval and fuzzy interval, marks the music scene the user is in based on his current state, and recommends corresponding personalized sports playlists, personalized leisure playlists, and personalized driving playlists according to the current music scene.

[0010] Furthermore, the state recognition unit monitors the user's heart rate through the monitoring terminal and constructs an effective heart rate graph based on the monitoring terminal, and the construction process is as follows:

[0011] When the user starts to use the monitoring terminal, the user's real-time heart rate is obtained and recorded as the monitored heart rate. The monitored heart rate within 24 hours of the user's first use is marked as the initial heart rate. A line graph of the initial heart rate over time is drawn and recorded as the heart rate change graph. A normal heart rate range interval is preset. ], remove the heart rate points outside the normal heart rate range in the heart rate variation graph, and record the removed heart rate data as abnormal data;

[0012] Calculate the ratio of abnormal data to total heart rate data. When the ratio is greater than or equal to a preset ratio threshold, generate an abnormal heart rate signal. Otherwise, record the heart rate change graph after removing the abnormal data as a valid heart rate graph.

[0013] Get the minimum and maximum heart rate values ​​in the effective heart rate graph and record them as valley heart rate. , peak heart rate , calculate the average heart rate in the effective heart rate graph to get the average heart rate , substitute into the formula The corrected mean value is calculated , the average value will be corrected Substitute into the formula , Calculate the resting interval ( ] and exercise range[ ),in These are all preset coefficient factors.

[0014] Furthermore, the state recognition unit obtains the acceleration value of the monitoring terminal and determines the current state of the user in combination with the resting interval and the exercise interval, wherein:

[0015] The acceleration of the monitoring terminal in three directions is continuously acquired through a three-axis acceleration sensor, the acceleration vector sum of the three directions is calculated, and the modulus of the summed vector is calculated to obtain the total acceleration, and a dynamic threshold and a static threshold are set, and the static threshold is smaller than the dynamic threshold;

[0016] When the user's current heart rate is in the exercise range and the total acceleration measured by the monitoring terminal is greater than or equal to the dynamic threshold, the user state is marked as exercise state;

[0017] When the user's current heart rate is in the resting range and the total acceleration measured by the monitoring terminal is less than or equal to the dynamic threshold, the user state is marked as resting;

[0018] Get the playback carrier currently bound to the playback terminal. When the playback carrier is the car computer, mark the user as driving state;

[0019] Among them, the marking priority of the driving state is higher than that of the moving state and the resting state.

[0020] Furthermore, the process of dividing the behavior intervals by the habit analysis unit is as follows:

[0021] Obtain the start time and end time of the user's driving state, and use the start time and end time of the driving state as interval endpoints to construct a driving interval;

[0022] Set an analysis duration and perform analysis according to the analysis duration interval. The analysis process is as follows:

[0023] Draw a line graph of the total acceleration change as an acceleration change graph, record the line portion above the dynamic threshold straight line as the dynamic portion, and record the interval portion above the static threshold straight line as the static portion;

[0024] The ratio of the projection length of the dynamic part of the acceleration change diagram on the time axis to the analysis time is calculated and recorded as the dynamic proportion. The proportion of the duration of the motion state in the analysis time is calculated and recorded as the exercise heart rate proportion. A first proportion threshold is set. When the exercise heart rate proportion and the dynamic proportion are both greater than or equal to the first proportion threshold, the time interval corresponding to the analysis time is marked as the exercise interval. When any multiple exercise intervals are continuous, they are merged and recorded as the same exercise interval.

[0025] The ratio of the projection length of the static part of the acceleration change diagram on the time axis to the analysis time is calculated and recorded as the static proportion. The proportion of the duration of the resting state in the analysis time is calculated and recorded as the resting heart rate proportion. A second proportion threshold is set. When the resting heart rate proportion and the static proportion are both greater than or equal to the second proportion threshold, the time interval corresponding to the analysis time is marked as the resting interval. When any multiple resting intervals are consecutive, they are merged and recorded as the same resting interval.

[0026] Furthermore, the process of constructing the user behavior habit distribution graph and behavior habit interval by the habit analysis unit is as follows:

[0027] The same behavior intervals were analyzed independently, and the analyzed behavior intervals were recorded as target intervals, and the corresponding behaviors were recorded as target behaviors;

[0028] Draw a target interval distribution map of the user every day, set a minimum number of analysis days N, obtain the target interval distribution maps of the user in the past N days and overlap them to obtain the user's 24-hour target behavior distribution overlap map, the target behavior distribution overlap map contains multiple overlapping intervals, assign an overlap value to each overlapping interval, the overlap value is equal to the number of target intervals containing the overlapping interval, calculate each overlapping value divided by the minimum number of analysis days to obtain a habit judgment value, when the habit judgment value is greater than the habit judgment threshold, record the overlapping interval corresponding to the overlapping value as the standard interval of the target behavior, obtain all overlapping intervals containing the standard interval and take the union to obtain a broad interval, and record the non-standard interval section of the broad interval as the fuzzy interval of the target behavior;

[0029] Mark the standard interval and fuzzy interval corresponding to each target behavior on the timeline to obtain the user's behavior habit distribution map.

[0030] Further, the process of the scene determination unit determining the current music scene is as follows:

[0031] When the opening moment occurs, obtain the standard interval or fuzzy interval in which the opening moment occurs;

[0032] When the opening time is within the standard interval, the target behavior corresponding to the standard interval is obtained, and the user's current music scene is marked as the music scene corresponding to the target behavior;

[0033] When the opening moment is in a fuzzy interval, the user's current user state is obtained, and when the target behavior corresponding to the fuzzy interval corresponds to the user's current user state, the user's current music scene is marked as the music scene corresponding to the target behavior.

[0034] Furthermore, it also includes a playlist construction unit, which obtains the user's favorite playlist and the preset playlist of the playback terminal for comprehensive analysis. The preset playlist includes a preset sports playlist, a preset leisure playlist and a preset driving playlist, thereby constructing a user's personalized recommended playlist.

[0035] Furthermore, the song list construction unit constructs the song list in the following process:

[0036] A music style database is preset, which contains a variety of song styles, and one or more song styles are marked for each song;

[0037] Get the song style composition in each preset playlist, where:

[0038] Record the songs in the preset song list of the playback terminal as preset songs, obtain the song styles of all the preset songs in the preset song list and form a set of song style categories, and count the number of times each song style appears in the preset song list as the song style repetition coefficient;

[0039] The songs in the user's favorite playlist are recorded as favorite songs, and the favorite songs are divided into playlists according to the style of each favorite song, wherein:

[0040] Compare the favorite songs with each preset playlist one by one, record the preset playlist to be compared as the comparison playlist, and obtain all the song styles of the favorite songs as , i is the number of the song style, substitute it into the formula Calculate the matching index ,in Indicates the style of the song The repetition coefficient in the preset song list. Same song style, =0;

[0041] Get all the repetition coefficients in each preset playlist and take the median as , get the total number of elements in the genre set of each preset playlist and record it as , get the total number of songs in each preset playlist and record it as , substitute into the formula Calculate in to get the matching threshold of the preset playlist ,in is the preset weight coefficient;

[0042] When the matching index between the favorite song and the comparison playlist is greater than the matching threshold of the comparison playlist, the favorite song is added to the preset playlist and constitutes a personalized recommended playlist.

[0043] Compared with the known prior art, the technical solution provided by the present invention has the following beneficial effects:

[0044] 1. The present invention analyzes and calculates the habit judgment value of a certain behavior of the user. When the habit judgment value is greater than the habit judgment threshold, it means that the user has performed the same target behavior at the same time for many times in the past N days, thereby judging and identifying the rules followed by various behaviors of the user and establishing the standard interval and fuzzy interval of the relevant behavior. When a target behavior has a standard interval, it is determined that the user will perform the target behavior that has become a habit within the time interval, while the fuzzy interval makes a larger range of inferences about the user's behavior, which makes up for the defect that the standard interval time range is too small and is not conducive to behavior analysis. Further verification based on the fuzzy interval can more efficiently judge the user's music scene

[0045] 2. The present invention determines whether the user is currently in a habitual behavior scene by analyzing and comparing the time interval when the user opens the app, thereby inferring the music scene the user is in. In addition, when in a fuzzy interval, the status verification is added to improve the accuracy of scene analysis when it is uncertain whether the user is in a habitual behavior scene.

[0046] 3. The present invention analyzes the style composition of a preset music scene playlist and then analyzes the songs collected by the user, thereby intelligently adding the user's favorite music to the preset playlist, so that the music scene playlist can be more in line with the user's daily preferences and enhance the user's listening experience. This makes up for the deficiency that the music scene playlist in the prior art is generally composed of pre-set songs and lacks personalized synchronous addition, and the user does not need to manually add songs to the music scene playlist to form a personalized playlist. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 It is the overall module block diagram of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] The present invention will be further described below in conjunction with the embodiments.

[0051] See also Figure 1 , a music scene recognition system based on artificial intelligence, at least includes: a monitoring terminal (including but not limited to smart bracelets, smart watches, smart rings, etc.) for monitoring dynamic data such as human heart rate, blood pressure, blood oxygen, positioning, etc., and a playback terminal for playing music. The monitoring terminal and the playback terminal use a wireless connection (such as Bluetooth) for data exchange, and also include: a state recognition unit, a habit analysis unit, and a scene determination unit.

[0052] The state recognition unit is used to identify the user's physical state and mark the user as being in motion, resting or driving state based on multiple dynamic data of the user's body.

[0053] When the user activates the monitoring terminal and starts using it, the monitoring terminal continuously obtains the user's real-time heart rate and records it as the monitored heart rate (the monitored heart rate is used to estimate the user's heart rate and can generally be regarded as the user's actual heart rate). The monitored heart rate within 24 hours of the user's first use is marked as the initial heart rate, and a line graph of the initial heart rate over time is drawn as the heart rate change graph. A normal heart rate range interval is preset [ ] (i.e., the range of human heart rate variation under normal circumstances, which is 40-220 in a specific embodiment), the heart rate points outside the normal heart rate range in the heart rate variation graph are removed, and the removed heart rate data are recorded as abnormal data;

[0054] Calculate the ratio of abnormal data to total heart rate data (the ratio of abnormal heart rate points to total heart rate points, i.e. the ratio of the duration of abnormal heart rate within 24 hours). When the ratio is greater than or equal to the preset ratio threshold, generate an abnormal heart rate signal. Otherwise, record the heart rate variation graph after removing the abnormal data as a valid heart rate graph. By screening the heart rate variation graph, avoid the abnormal data in the heart rate variation graph from interfering with the analysis results.

[0055] Get the minimum and maximum heart rate values ​​in the effective heart rate graph and record them as valley heart rate. , peak heart rate , calculate the average heart rate in the effective heart rate graph to get the average heart rate , substitute into the formula The corrected mean value is calculated (Reduce the impact of peak heart rate, so as to obtain a corrected average value that is more in line with the user's average heart rate during daily activities), Substitute into the formula , Calculate the resting interval ( ] and exercise range[ ),in All are preset coefficient factors;

[0056] By constructing an effective heart rate graph, we can obtain the valley heart rate, peak heart rate and average heart rate that can reflect the user's heart rate level. The speed of the heart rate can indirectly reflect the user's physical fitness. Generally speaking, the lower the resting heart rate and the slower the average heart rate, the stronger the user's heart pumping ability and the stronger the myocardial function, the better the corresponding physical fitness and the higher the maximum achievable heart rate. Therefore, by analyzing the user's initial heart rate, we can quickly know the user's physical fitness and judge the user's physical state by setting the corresponding interval. Compared with fixed intervals, it is more suitable for people with different physical fitness and improves personalized state analysis.

[0057] The acceleration of the monitoring terminal in three directions is continuously acquired through a three-axis acceleration sensor, the acceleration vector sum of the three directions is calculated, and the modulus of the summed vector is calculated to obtain the total acceleration, and a dynamic threshold and a static threshold are set, and the static threshold is smaller than the dynamic threshold;

[0058] When the user's current heart rate is in the exercise range and the total acceleration measured by the monitoring terminal is greater than or equal to the dynamic threshold, the user state is marked as exercise state;

[0059] When the user's current heart rate is in the resting range and the total acceleration measured by the monitoring terminal is less than or equal to the dynamic threshold, the user state is marked as resting;

[0060] It should be noted that judging the user's state by monitoring the user's heart rate and comprehensively considering the acceleration change caused by the user's activities can avoid the influence of the user's emotional fluctuations on the judgment of the exercise state, compared with judging whether the user is moving or still by only one of the heart rate or acceleration, and thus make the judgment result more accurate.

[0061] Get the playback carrier currently bound to the playback terminal (playback carriers include speakers, mobile phones, car computers, etc.). When the playback carrier is the car computer, generate a mark that the user is in driving state;

[0062] Among them, the marking priority of the driving state is higher than the moving state and the resting state. That is to say, when the driving state and other state marks exist at the same time, the driving state is used as the current state.

[0063] Furthermore, the scene recognition unit analyzes the user's historical status data and defines three behavior intervals of the user, namely, a driving interval, a sports interval, and a resting interval, and the corresponding behaviors are driving, sports, and resting, respectively. The user's behavior habits are determined according to the distribution of the user's behavior intervals, thereby constructing a user behavior habit interval distribution map;

[0064] The behavior interval division process is as follows:

[0065] Obtain the start time and end time of the user's driving state, and use the start time and end time of the driving state as interval endpoints to construct a driving interval;

[0066] An analysis time is set (in a specific embodiment, the analysis time is 10 minutes), and analysis is performed once every analysis time. The analysis process is as follows:

[0067] Draw a line graph of the total acceleration of the monitoring terminal changing with time as an acceleration change graph, record the part of the total acceleration line in the acceleration change graph that is above the dynamic threshold straight line as a dynamic part, and record the part of the total acceleration line in the acceleration change graph that is above the static threshold straight line as a static part;

[0068] The ratio of the projection length of the dynamic part of the acceleration change diagram on the time axis to the analysis time is calculated and recorded as the dynamic proportion. The proportion of the duration of the motion state in the analysis time is calculated and recorded as the exercise heart rate proportion. A first proportion threshold is set. When the exercise heart rate proportion and the dynamic proportion are both greater than or equal to the first proportion threshold, the time interval corresponding to the analysis time is marked as the exercise interval. When any multiple exercise intervals are continuous, they are merged and recorded as the same exercise interval.

[0069] The ratio of the projection length of the static part of the acceleration change diagram on the time axis to the analysis time is calculated and recorded as the static proportion. The proportion of the duration of the resting state in the analysis time is calculated and recorded as the resting heart rate proportion. A second proportion threshold is set. When the resting heart rate proportion and the static proportion are both greater than or equal to the second proportion threshold, the time interval corresponding to the analysis time is marked as the resting interval. When any multiple resting intervals are consecutive, they are merged and recorded as the same resting interval.

[0070] The process of constructing the user behavior habit interval is as follows:

[0071] The same behavior intervals were analyzed independently, and the analyzed behavior intervals were recorded as target intervals, and the corresponding behaviors were recorded as target behaviors;

[0072] Draw a target interval distribution map of the user every day (construct a timeline of the day, and use a dark area to mark the target interval on the timeline), set a minimum number of analysis days N (in a specific embodiment, N is 21), obtain the target interval distribution map of the user in the past N days and overlap them to obtain the user's 24-hour target behavior distribution overlap map, the target behavior distribution overlap map contains multiple overlapping intervals, and assign an overlap value to each overlapping interval, the overlap value is equal to the number of target intervals containing the overlapping interval (when both endpoints of an overlapping interval are located in a target interval, the target interval contains the overlapping interval), calculate each overlap value divided by the minimum number of analysis days to obtain a habit judgment value, when the habit judgment value is greater than the habit judgment threshold, the overlapping interval corresponding to the overlap value is recorded as the standard interval of the target behavior, obtain all overlapping intervals containing the standard interval and take the union to obtain a broad interval, and record the interval part of the non-standard interval in the broad interval as the fuzzy interval of the target behavior;

[0073] It should be noted that when the habit judgment value is greater than the habit judgment threshold, it means that the user has performed the same target behavior at the same time many times in the past N days. Generally, the user's various behaviors will follow certain rules, such as driving, exercising or resting at the same time every day. Therefore, when a target behavior has a standard interval, it can be determined that the user will perform the habitual target behavior within the time interval. The fuzzy interval speculates on the user's behavior in a larger range, which makes up for the defect that the standard interval time range is too small to be conducive to behavior analysis. Further verification based on the fuzzy interval can more efficiently judge the user's music scene.

[0074] The standard interval and fuzzy interval corresponding to each target behavior are marked on the same 24-hour timeline. Each standard interval or fuzzy interval corresponds to a target behavior. That is, a corresponding target behavior (driving, exercise, rest) can be determined through the interval (standard interval and fuzzy interval) in which the current time is located, and then the user's current music scene can be further marked according to the target behavior to obtain the user's behavior habit distribution map.

[0075] The scene determination unit is used to determine the user's current music scene, which includes sports scenes, rest scenes or driving scenes. The three music scenes, target behaviors and user states correspond to each other (for example, sports corresponds to sports scenes and sports states), and recommends corresponding personalized sports playlists, personalized rest playlists and personalized driving playlists based on the current music scene.

[0076] The music scene determination process is specifically as follows:

[0077] When the user turns on the playback terminal, the moment is recorded as the opening moment, and the standard interval or fuzzy interval where the opening moment is located is obtained;

[0078] When the opening time is within the standard interval, the target behavior corresponding to the standard interval is obtained, and the user's current music scene is marked as the music scene corresponding to the target behavior (for example, when the opening time is within the standard interval corresponding to the driving behavior, the user's music scene is marked as the driving scene);

[0079] When the opening moment is in a fuzzy interval, the user's current user state is obtained. When the target behavior corresponding to the fuzzy interval corresponds to the user's current user state, the user's current music scene is marked as the music scene corresponding to the target behavior (for example, when the opening moment is in a fuzzy interval corresponding to the driving behavior, the user's current state is obtained. If the user's current state is a driving state, the user's music scene is marked as a driving scene).

[0080] It should be noted that by analyzing and comparing the time interval when the user opens the app, we can determine whether the user is currently in a habitual behavior scenario, thereby inferring the music scene the user is in. In addition, when in a fuzzy interval, the status verification is added to improve the accuracy of scene analysis when it is uncertain whether the user is in a habitual behavior scenario.

[0081] It also includes a playlist construction unit, which performs a comprehensive analysis on the user's favorite playlist (the favorite playlist consists of songs collected and subscribed by the user, that is, a playlist created by the user through the playback terminal and including the songs he likes to listen to) and the preset playlist of the playback terminal. The preset playlist includes a preset sports playlist, a preset leisure playlist and a preset driving playlist, thereby constructing a personalized recommended playlist for the user.

[0082] A music style database is preset, which contains a variety of song styles, and one or more song styles are marked for each song (the determination of song styles is a prior art, and each song will be marked with a song style when it is released, which will not be described in detail here);

[0083] Get the song style composition in each preset playlist, where:

[0084] Record the songs in the preset song list of the playback terminal as preset songs, obtain the song styles of all the preset songs in the preset song list and form a music style category set (the music style category set includes the song styles of all the preset songs, and the song styles are not repeated), count the number of times various song styles appear in the preset song list and record it as the song style repetition coefficient, thereby obtaining a music style category set containing multiple song styles and the repetition coefficient of each song style;

[0085] The songs in the user's favorite playlist are recorded as favorite songs, and the favorite songs are divided into playlists according to the style of each favorite song, wherein:

[0086] Compare the favorite songs with each preset playlist one by one, record the preset playlist to be compared as the comparison playlist, and obtain all the song styles of the favorite songs as , i is the number of the song style, substitute it into the formula Calculate the matching index The matching index is used to indicate the similarity between the style of the collection song and the preset playlist, so as to determine the applicable music scene of the collection song. Indicates the style of the song The repetition coefficient in the preset song list. Same song style, =0;

[0087] Get all the repetition coefficients in each preset playlist and take the median as , get the total number of elements in the genre set of each preset playlist and record it as , get the total number of songs in each preset playlist and record it as , substitute into the formula Calculate in to get the matching threshold of the preset playlist ,in is the preset weight coefficient;

[0088] When the matching index between the favorite song and the comparison playlist is greater than the matching threshold of the comparison playlist, the favorite song is added to the preset playlist and forms a personalized recommended playlist (when the favorite song already exists in the preset playlist, the favorite song will not be divided).

[0089] It should be noted that the music scene playlist in the prior art is generally composed of pre-set songs and lacks personalized synchronous addition. Users need to manually add songs to the music scene playlist to build a personalized playlist. The playlist construction unit analyzes the preset music scene playlist style composition and then analyzes the user's favorite songs, thereby intelligently adding the user's favorite music to the preset playlist, so that the music scene playlist can be more in line with the user's daily preferences and enhance the user's listening experience.

[0090] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A music scene recognition system based on artificial intelligence, comprising a monitoring terminal for monitoring human body data and a playing terminal for playing music, characterized in that: Also includes: A state recognition unit obtains and analyzes the user's body data and marks the user as being in motion, resting, or driving state; The habit analysis unit obtains the user's historical status data and combines the acceleration change curve of the monitoring terminal to define the behavior intervals, including driving intervals, exercise intervals and rest intervals. The corresponding behaviors are driving, exercise and rest. The user's historical behavior intervals are obtained and overlap analysis is performed to clarify the overlap intervals of each behavior and assign overlap values. The overlap intervals with overlap values ​​greater than the preset threshold are selected as standard intervals, and fuzzy intervals are constructed based on the overlap intervals and standard intervals. The habit analysis unit analyzes the behavior interval distribution, thereby establishing a standard interval and a fuzzy interval, and constructing a user behavior habit distribution map, the process is as follows: Conduct independent analysis on the same behavior interval, record the analyzed behavior interval as the target interval, and record the corresponding behavior as the target behavior; Draw a target interval distribution map of the user every day, set a minimum number of analysis days N, obtain the target interval distribution maps of the user in the past N days and overlap them to obtain the user's 24-hour target behavior distribution overlap map, the target behavior distribution overlap map contains multiple overlapping intervals, assign an overlap value to each overlapping interval, the overlap value is equal to the number of target intervals containing the overlapping interval, calculate each overlapping value divided by the minimum number of analysis days to obtain a habit judgment value, when the habit judgment value is greater than the habit judgment threshold, record the overlapping interval corresponding to the overlapping value as the standard interval of the target behavior, obtain all overlapping intervals containing the standard interval and take the union to obtain a broad interval, and record the non-standard interval section of the broad interval as the fuzzy interval of the target behavior; Mark the standard interval and fuzzy interval corresponding to each target behavior on the time axis to obtain the user's behavior habit distribution map; The scene determination unit determines the opening time of the playback terminal, compares the opening time with the standard intervals and fuzzy intervals, marks the music scene the user is in based on his current status, and recommends corresponding personalized sports playlists, personalized leisure playlists, and personalized driving playlists based on the current music scene.

2. The music scene recognition system based on artificial intelligence according to claim 1, characterized in that: The state recognition unit monitors the user's heart rate through the monitoring terminal and constructs an effective heart rate graph based on it. The construction process is as follows: When the user starts to use the monitoring terminal, the user's real-time heart rate is obtained and recorded as the monitored heart rate. The monitored heart rate within 24 hours of the user's first use is marked as the initial heart rate. A line graph of the initial heart rate over time is drawn and recorded as the heart rate change graph. A normal heart rate range interval is preset. ], remove the heart rate points outside the normal heart rate range in the heart rate variation graph, and record the removed heart rate data as abnormal data; Calculate the ratio of abnormal data to total heart rate data. When the ratio is greater than or equal to a preset ratio threshold, generate an abnormal heart rate signal. Otherwise, record the heart rate change graph after removing the abnormal data as a valid heart rate graph. Get the minimum and maximum heart rate values ​​in the effective heart rate graph and record them as valley heart rate. , peak heart rate , calculate the average heart rate in the effective heart rate graph to get the average heart rate , substitute into the formula The corrected mean value is calculated , the average value will be corrected Substitute into the formula , Calculate the resting interval ( ] and exercise range[ ),in These are all preset coefficient factors.

3. The music scene recognition system based on artificial intelligence according to claim 2, characterized in that: The state recognition unit obtains the acceleration value of the monitoring terminal and determines the current state of the user in combination with the resting interval and the moving interval, wherein: The acceleration of the monitoring terminal in three directions is continuously acquired through a three-axis acceleration sensor, the acceleration vector sum of the three directions is calculated, and the modulus of the summed vector is calculated to obtain the total acceleration, and a dynamic threshold and a static threshold are set, and the static threshold is smaller than the dynamic threshold; When the user's current heart rate is in the exercise range and the total acceleration measured by the monitoring terminal is greater than or equal to the dynamic threshold, the user state is marked as exercise state; When the user's current heart rate is in the resting range and the total acceleration measured by the monitoring terminal is less than or equal to the dynamic threshold, the user state is marked as resting; Get the playback carrier currently bound to the playback terminal. When the playback carrier is the car computer, mark the user as driving state; Among them, the marking priority of the driving state is higher than that of the moving state and the resting state.

4. The music scene recognition system based on artificial intelligence according to claim 1, characterized in that: The process of dividing the behavior intervals by the habit analysis unit is as follows: Obtain the start time and end time of the user's driving state, and use the start time and end time of the driving state as interval endpoints to construct a driving interval; Set an analysis duration and perform analysis according to the analysis duration interval. The analysis process is as follows: Draw a line graph of the total acceleration change as an acceleration change graph, record the line portion above the dynamic threshold straight line as the dynamic portion, and record the interval portion above the static threshold straight line as the static portion; The ratio of the projection length of the dynamic part of the acceleration change diagram on the time axis to the analysis time is calculated and recorded as the dynamic proportion. The proportion of the duration of the motion state in the analysis time is calculated and recorded as the exercise heart rate proportion. The exercise heart rate proportion and the dynamic proportion are compared with the first proportion threshold to determine whether the time interval corresponding to the analysis time is marked as the exercise interval. When any multiple exercise intervals are continuous, they are merged and recorded as the same exercise interval. The ratio of the projection length of the static part of the acceleration change diagram on the time axis to the analysis time is calculated and recorded as the static proportion. The proportion of the duration of the resting state in the analysis time is calculated and recorded as the resting heart rate proportion. The resting heart rate proportion and the static proportion are compared with the second proportion threshold to determine whether the time interval corresponding to the analysis time is marked as a resting interval. When any multiple resting intervals are consecutive, they are merged and recorded as the same resting interval.

5. The music scene recognition system based on artificial intelligence according to claim 4, characterized in that: The process of determining the current music scene by the scene determination unit is as follows: When the opening moment occurs, obtain the standard interval or fuzzy interval in which the opening moment occurs; When the opening time is within the standard interval, the target behavior corresponding to the standard interval is obtained, and the user's current music scene is marked as the music scene corresponding to the target behavior; When the opening moment is in a fuzzy interval, the user's current user state is obtained, and when the target behavior corresponding to the fuzzy interval corresponds to the user's current user state, the user's current music scene is marked as the music scene corresponding to the target behavior.

6. The music scene recognition system based on artificial intelligence according to claim 1, characterized in that: It also includes a playlist construction unit, which obtains the user's favorite playlist and the preset playlist of the playback terminal for comprehensive analysis. The preset playlist includes a preset sports playlist, a preset leisure playlist and a preset driving playlist, thereby constructing a personalized recommended playlist for the user.

7. The music scene recognition system based on artificial intelligence according to claim 6, characterized in that: The process of constructing a song list by the song list construction unit is as follows: A music style database is preset, which contains a variety of song styles, and one or more song styles are marked for each song; Get the song style composition in each preset playlist, where: Record the songs in the preset song list of the playback terminal as preset songs, obtain the song styles of all the preset songs in the preset song list and form a set of song style categories, and count the number of times each song style appears in the preset song list as the song style repetition coefficient; The songs in the user's favorite playlist are recorded as favorite songs, and the favorite songs are divided into playlists according to the style of each favorite song, wherein: Compare the favorite songs with each preset playlist one by one, record the preset playlist to be compared as the comparison playlist, and obtain all the song styles of the favorite songs as , i is the number of the song style, substitute it into the formula Calculate the matching index ,in Indicates the style of the song The repetition coefficient in the preset song list. Same song style, =0; Get all the repetition coefficients in each preset playlist and take the median as , get the total number of elements in the genre set of each preset playlist and record it as , get the total number of songs in each preset playlist and record it as , substitute into the formula Calculate in to get the matching threshold of the preset playlist ,in is the preset weight coefficient; When the matching index between the favorite song and the comparison playlist is greater than the matching threshold of the comparison playlist, the favorite song is added to the preset playlist and constitutes a personalized recommended playlist.

Images