A Smart Music Recommendation Method and System for Elderly People Living at Home

Abstract

This invention discloses an intelligent music recommendation method and system for elderly people living at home, aiming to solve the technical problems of existing music recommendation systems neglecting the physiological characteristics, listening time patterns, and gender differences of elderly users, resulting in low recommendation accuracy and poor user experience. The system includes a data acquisition module, an elderly person profile construction module, a song feature extraction module, a spatiotemporal matching engine, and a push interface. The data acquisition module obtains the gender information, listening timestamps, and playback records of elderly people living at home; the elderly person profile construction module classifies elderly people into types based on gender and age, and presets corresponding physiological rhythm weight coefficients for each type; the song feature extraction module analyzes the BPM, pitch, emotional polarity, and genre characteristics of music files; the spatiotemporal matching engine retrieves target feature intervals from a preset mapping table based on listening time and elderly person type, calculates the matching degree with song features, and sorts them; the push interface sends the sorted song list to the playback terminal in the elderly person's room. This invention, by introducing a gender dimension and a listening time weighting mechanism, combined with a time-period awareness algorithm, achieves precise and personalized music recommendations for elderly people living at home, effectively improving the listening experience for elderly users.

Description

TECHNICAL FIELD

[0002] The present application relates to the field of artificial intelligence and big data processing technology, in particular to an intelligent music recommendation method and system capable of multi-dimensional fusion analysis according to the gender of the home elderly, the listening time period, the listening time length and the song type characteristics, and dynamic adjustment of the push strategy. BACKGROUND

[0003] With the intensification of population aging trend, the demand for intelligent services in the home care scene is growing. Music, as an important spiritual comfort means in the daily life of the elderly, the accuracy and timeliness of its push directly affect the quality of life of the elderly.

[0004] The existing music recommendation system (such as collaborative filtering, content-based recommendation) mainly has the following technical problems: (1) Static portrait limitation: only relying on the historical listening preferences of users or demographic information, ignoring the dynamic change needs of elderly users in different life stages. For example, male elderly users may prefer high-pitched music such as operas and operas, while female elderly users prefer soothing music such as folk songs and light music. The traditional system cannot automatically adjust the recommendation weight according to gender.

[0005] (2) Ignoring time context: existing daily recommendations or playlist recommendations fail to distinguish specific listening times. For example, fast-paced music is needed to refresh in the morning commute period, while slow music is needed to help sleep at night. If the same content is pushed all day, it will seriously affect the elderly user's work and rest schedule.

[0006] (3) Ignoring listening time length signal: traditional systems only record the number of plays, without distinguishing between complete plays and fragment plays. Elderly users may generate a large number of false touch short play records due to inexperience, and these noise data will seriously pollute the user portrait, leading to a deviation of the recommended preferences from the true preferences.

[0007] (4) Content and emotion are separated: the system often only recommends according to the genre tag, lacking analysis of deep content features such as song BPM (beats per minute), pitch, and emotional polarity, resulting in recommended songs that meet user preferences but do not match the current physiological state (such as heart rate, fatigue).

[0008] Therefore, there is an urgent need for an intelligent push system that can comprehensively consider the gender characteristics of home elderly, real-time time dimension, listening time length quality, and deep content features of songs. SUMMARY

[0009] The purpose of this invention is to provide a smart music recommendation method and system for elderly people living at home. By constructing a multi-dimensional user profile and combining time-aware algorithms, gender-differentiated weights, and listening time quality assessment, the invention achieves accurate and personalized music recommendation.

[0010] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A smart music recommendation method includes a data collection step, a profile construction step, a song feature extraction step, a spatiotemporal matching step, and a recommendation step.

[0011] The data collection step is used to collect the registration information (including gender and age) of elderly people living at home, real-time listening timestamps, and historical playback records (including song type, playback start and end time, and effective playback duration).

[0012] The profile construction step is used to classify elderly people into different types (such as "active male elderly people" and "quiet female elderly people") based on their gender and age data, and to preset the corresponding physiological rhythm weight coefficients for each type.

[0013] The song feature extraction step is used to analyze the deep features of music files, including but not limited to BPM, pitch, genre, emotional polarity of lyrics, and instrument composition.

[0014] The spatiotemporal matching step is the core processing step, used to execute the following logic: receive the timestamp of the current music listening request and determine its time period; retrieve the gender type label from the elderly person's profile, and in combination with the current time period, retrieve the target feature interval from the preset "time period-gender-music feature mapping table"; calculate the similarity between the target feature interval and the song feature extraction results, and filter out the candidate song set with the highest matching degree; sort the candidate songs according to the matching degree.

[0015] The push step is used to send the sorted song list to the playback terminal in the elderly person's room.

[0016] Furthermore, the system also includes a feedback optimization module, which records the elderly’s actions on the pushed songs (such as skipping, favorite, repeating, and adjusting the volume), and dynamically adjusts the weight coefficients in the “time period-gender-music feature mapping table” based on the feedback data.

[0017] Furthermore, the rules for determining the effective playback duration are as follows: if the single playback duration exceeds 70% of the total song duration or the continuous playback exceeds 180 seconds, it is marked as a complete playback and assigned a weight of 1.0; if the single playback duration is less than 30 seconds, it is determined as a mis-touch operation and is not included in the profile statistics; if the single playback duration is between 30 seconds and 180 seconds, it is marked as a segment playback and assigned a linear decay weight of 0.3 to 0.8. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the architecture of the intelligent music push system of the present invention.

[0019] Figure 2 This is a flowchart of the intelligent music push method of the present invention.

[0020] Figure 3 This is a schematic diagram of the data structure of the "Time Period-Gender-Music Feature Mapping Table" in an embodiment of the present invention.

[0021] Figure 4 This is a flowchart illustrating the logic of weighted processing based on listening time in an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0023] like Figure 1 and Figure 2 As shown, this embodiment provides an intelligent music push system for elderly people living at home, which mainly includes a data acquisition module 100, an elderly person profile construction module 200, a song feature extraction module 300, a spatiotemporal matching engine 400, a push interface 500, and a feedback optimization module 600.

[0024] With the assistance of their children or community staff, elderly people living at home complete registration on the playback terminal 101 (such as a voice speaker with a large-font screen). The data collection module 100 obtains information such as the elderly person's gender (e.g., "female") and age (e.g., 72 years old). At the same time, the data collection module 100 monitors the elderly person's music listening behavior in real time, recording the precise timestamp of each request (e.g., 7:30 AM), song type tags (e.g., "folk song" "light music"), and the start and end times of playback.

[0025] Optionally, the system can connect to the wearable device 102 (such as a smart bracelet) worn by the elderly via Bluetooth to obtain the elderly's current real-time heart rate data as an auxiliary basis for judging physiological status.

[0026] The elderly profile building module 200 receives gender and age data, combines it with music listening behavior logs, and labels the elderly with different types of data. The system has pre-set physiological rhythm models for different genders and age groups. Elderly men (65-75 years old): In the morning, they prefer loud music such as marches and revolutionary songs (BPM 100-130); in the afternoon, they prefer audio of language such as opera and crosstalk; and at night, they prefer light music to help them sleep.

[0027] Elderly women (65-75 years old): Prefer folk songs and lyrical songs throughout the day (BPM 60-100), pay attention to beautiful melodies and emotional resonance, and have a habit of listening to health and wellness audio in the afternoon.

[0028] Senior citizens (75 years and older): Prefer nostalgic classics (BPM 50-80), have low volume requirements, and are less receptive to unfamiliar songs; the system automatically reduces the exploration rate.

[0029] Based on this, different types of elderly people are assigned differentiated "vitality weight" and "nostalgia weight".

[0030] This is one of the core innovations that distinguishes this invention from existing technologies. For example... Figure 4 As shown, when recording playback behavior, the song feature extraction module 300 not only records "whether it is playing", but also accurately calculates the effective playback duration and assigns quality weights according to the following rules: Rule A (Complete Playback): A single playback duration exceeding 70% of the song's total duration, or continuous playback exceeding 180 seconds, is marked as a complete playback, with a weight of 1.0.

[0031] Rule B (Segment Playback): Single playback duration between 30 and 180 seconds is marked as segment playback, with a weight of 0.3 to 0.8 that decreases linearly (the longer the playback, the higher the weight).

[0032] Rule C (Mistake Removal): If a single playback duration is less than 30 seconds, it is judged as a mistake (the elderly may press the wrong button) and is directly removed from the profile.

[0033] Through the above rules, the system effectively filters out noisy data generated by elderly users' unfamiliarity with the operation, making the profile closer to their real preferences.

[0034] The song feature extraction module 300 preprocesses the songs in the music library. For example, for song A (such as "Jasmine Flower"), its features are extracted as follows: BPM=72, mode=pentatonic, emotional polarity=positive, genre=folk song; for song B (such as "Returning from Target Practice"), its features are extracted as follows: BPM=128, mode=major, emotional polarity=positive, genre=march; for song C (such as "Moon Reflected in the Second Spring"), its features are extracted as follows: BPM=52, mode=minor, emotional polarity=neutral to melancholic, genre=folk music solo.

[0035] The spatiotemporal matching engine 400 executes the push logic: (1) Determine that the current time is the "early morning wake-up period" (6:00-9:00).

[0036] (2) Search the mapping table database 401 and find the rule: "early morning wake-up time + female elderly" corresponding recommended features: BPM 60-90, pentatonic scale / major key, positive emotion, folk song or light music type.

[0037] (3) Calculate the matching degree: The feature of song A (72BPM, pentatonic scale, positive, folk song) has a high degree of overlap with the target feature range of up to 92%; the feature of song B (128BPM, march) has an overlap of only 30%; the feature of song C (52BPM, minor key, melancholy) has an overlap of 40% but the emotional polarity does not match.

[0038] (4) Generate a recommendation sequence: put song A at the top of the recommendation list, song C next (although the BPM is low, it is a nostalgic classic as a backup), and song B at the bottom.

[0039] (5) If the wearable device 102 data is connected and the user’s current heart rate is detected to be low at rest, the engine 400 will further increase the recommendation weight of medium and high BPM songs to help the elderly quickly enter a state of wakefulness and activity.

[0040] The push interface 500 sends the playlist to the playback terminal 101 in the elderly person's room. The terminal has a large font display, physical knob volume control, one-button call for children, and supports short voice commands such as "play music," "stop," and "next song," reducing the learning curve for the elderly.

[0041] The feedback optimization module records 600 instances of elderly behavior: if an elderly person finishes playing song A and saves it, the weight of "early morning + female + folk song" is increased; if the elderly person immediately skips song B, the weight of "early morning + female + march" is decreased. After 2-4 weeks of online learning, the system's push accuracy can be improved to over 85%. Beneficial effects

[0042] 1. Gender-differentiated recommendation: By introducing a gender-dimensional preference vector, the problem of "not distinguishing between men and women" in traditional recommendation systems is solved, making recommendations more in line with the gender-specific aesthetic preferences of elderly users.

[0043] 2. Quality-weighted listening time: By distinguishing between full playback, segment playback, and accidental touch operations, noise data caused by elderly users' unfamiliarity with operation is effectively filtered out, significantly improving the accuracy of the user profile.

[0044] 3. Time-based recommendation: By dividing the time into five segments, the recommended music is synchronized with the physiological rhythm of the elderly, avoiding problems such as pushing fast songs late at night that affect sleep.

[0045] 4. Lowering the barrier to entry: The playback terminal is designed specifically for this system. The combination of large fonts, physical knobs, and voice commands allows elderly people who are not good at using smartphones to easily enjoy personalized music services.

[0046] 5. Continuous self-evolution: Through the feedback optimization module, the system dynamically adjusts the weights according to changes in the elderly's listening habits, achieving long-term adaptability that "understands you better the more you use it". Instruction manual illustrations

[0047] Figure 1 This diagram illustrates the overall architecture of the intelligent music recommendation system of this invention. In the diagram, 100 represents the data acquisition module, which acquires user data through the elderly terminal 101 (voice speaker) and wearable device 102 (smart bracelet). 200 represents the elderly profile construction module, which receives the output from the data acquisition module and generates profile data including gender preference vectors and time-period preference distributions. 300 represents the song feature extraction module, which connects to the music library and outputs feature vectors such as BPM, pitch, and emotional polarity for each song. 400 represents the spatiotemporal matching engine, which connects to the profile construction module 200 and the song feature extraction module 300, and has a built-in mapping table database 401 to execute the core matching algorithm. 500 represents the push interface, which sends the matching results to the elderly terminal 101. 600 represents the feedback optimization module, which collects the elderly's actions regarding the recommended songs and sends this data back to the profile construction module 200 and the spatiotemporal matching engine 400 for weight adjustment.

[0048] Solid arrows indicate the direction of data transmission, while dashed arrows indicate the direction of feedback adjustment.

[0049] Figure 2 This paper demonstrates the complete process of the intelligent music push method of the present invention. The process starts from the "start" node and proceeds in the following order: obtaining the gender and age information of the elderly (S201), collecting listening behavior logs including timestamps and playback durations (S202), dividing listening time periods (S203), marking complete / fragment / accidental playback and assigning weights (S204), constructing a gender dimension preference vector (S205), determining the current time period and retrieving the mapping table (S206), calculating the song feature matching degree (S207), sorting and generating a push list (S208), sending it to the playback terminal (S209), recording feedback behavior (S210), and determining whether the update conditions are met (S211). If the conditions are met, the process returns to S205 to update the profile weights; otherwise, it waits for the next push to be triggered.

[0050] Figure 3The core mapping relationship of this invention is presented in tabular form. The horizontal header of the table represents five listening periods (morning wake-up period, morning activity period, afternoon nap period, evening leisure period, and nighttime sleep aid period), and the vertical header represents four types of elderly people (active male elderly people, quiet male elderly people, active female elderly people, and quiet female elderly people). Each cell contains the target BPM range, recommended key, emotional polarity preference, and type priority corresponding to the "period + type" combination. For example, the cell for "morning wake-up period + active male elderly people" contains: BPM 110-140, major key, positive, march > revolutionary songs > folk music; the cell for "nighttime sleep aid period + quiet female elderly people" contains: BPM 40-65, minor key or pentatonic scale, neutral to positive, light music > folk slow songs > white noise.

[0051] Figure 4 This demonstrates the specific logic of the music listening time quality assessment of this invention. The process begins with "recording a playback action" (S401), obtains the actual playback duration T (S402), and determines whether T is greater than 70% of the total song duration or T>180 seconds (S403). If so, it is marked as a complete playback, with a weight W=1.0 (S404). If not, it determines whether T is less than 30 seconds (S405). If so, it is determined as a false touch, W=0, and it is not included in the profile (S406). If not (30 seconds<=T<=180 seconds), the weight is calculated according to the linear decay formula W=0.3+0.5*(T-30) / (180-30) (S407), and the final weight value W is output and used in the profile statistics (S408).

Claims

1. A smart music recommendation method for elderly people living at home, characterized in that, Includes the following steps: Obtain user registration attributes for elderly people living at home, including at least gender and age; Collect the elderly person's music listening behavior log within a preset history window. Each log entry records the song's type tag, playback start timestamp, playback end timestamp, and effective playback duration. The listening time periods are divided according to the playback start timestamp, and the listening time periods include at least the morning wake-up period, the morning activity period, the lunch break period, the evening leisure period, and the nighttime sleep aid period. Based on the relationship between the effective playback duration and the preset duration threshold, the playback is marked as a complete playback or a segment playback, and segment playback is assigned a lower weight than complete playback. Construct a music profile of the elderly, which includes: a gender-dimensional preference vector, a genre preference distribution under each listening time period, and an average effective playback duration for each time period; In response to the current push notification event, the current time period is determined, and the target type distribution and rhythm feature interval of the corresponding time period are retrieved from the elderly music profile. The target song feature vector is obtained by weighted fusion with the gender dimension preference vector. Search the music library for the set of songs that have the highest matching degree with the feature vector of the target song, and generate a push list; Playback is performed via a playback terminal deployed in the elderly person's room.

2. The intelligent music recommendation method according to claim 1, characterized in that, The specific time boundaries for listening to music are as follows: the morning wake-up period is 6:00-9:00, the morning activity period is 9:00-12:00, the afternoon nap period is 12:00-14:00, the evening leisure period is 17:00-20:00, and the nighttime sleep aid period is 20:00-23:

00.

3. The intelligent music push method according to claim 1, characterized in that, The gender dimension preference vector is constructed as follows: for elderly men, the weight coefficients of uplifting, marching, and opera songs are increased; for elderly women, the weight coefficients of lyrical, folk, and light music songs are increased.

4. The intelligent music recommendation method according to claim 1, characterized in that, The rules for determining the effective playback duration are as follows: if the single playback duration exceeds 70% of the total song duration or the continuous playback exceeds 180 seconds, it is marked as a complete playback and assigned a weight of 1.0; if the single playback duration is less than 30 seconds, it is judged as a mis-touch operation and is not included in the profile statistics; if the single playback duration is between 30 seconds and 180 seconds, it is marked as a segment playback and assigned a linear decay weight of 0.3 to 0.

8.

5. The intelligent music push method according to claim 1, characterized in that, The push trigger event includes at least one of the following: the playback terminal detects the elderly person's voice command to sing, the timer reaches the preset time period trigger point, and the wearable device detects that the elderly person's heart rate is lower than the resting threshold for more than 10 minutes.

6. The intelligent music recommendation method according to claim 1, characterized in that, The playback terminal is a wall-mounted audio device or a voice speaker with a physical knob. The playback terminal is equipped with a large-font display screen and a physical button for one-click calling of children.

7. The intelligent music recommendation method according to claim 1, characterized in that, The matching degree calculation adopts the cosine similarity algorithm and introduces a time period attenuation coefficient. The time period attenuation coefficient is dynamically adjusted according to the elderly's listening frequency in the past 7 days. The higher the listening frequency, the smaller the attenuation coefficient.

8. The intelligent music push method according to claim 1, characterized in that, It also includes feedback optimization steps: recording the elderly’s actions on the pushed songs, including skipping, favorite, repeating, and adjusting the volume, and dynamically adjusting the weight coefficients in the gender dimension preference vector and time period type preference distribution based on the feedback data.

9. A smart music recommendation system for elderly people living at home, characterized in that, include: The data acquisition module is used to obtain the gender information, listening timestamps, and playback records of elderly people living at home. The elderly profile construction module is used to construct an elderly music profile based on gender and age. The profile includes a gender-dimensional preference vector and a type preference distribution under each listening period. The song feature extraction module is used to analyze the song's BPM, pitch, emotional polarity, and genre characteristics. The spatiotemporal matching engine is used to retrieve target feature intervals from a preset mapping table based on the current listening time and the type of elderly person, and calculate the matching degree with the song features; The push interface is used to send the matched song list to the playback terminal in the elderly person's room.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 8.

Images