CN119632550B - A multifunctional system based on a single accelerometer - Google Patents
A multifunctional system based on a single accelerometerInfo
- Publication number
- CN119632550B CN119632550B CN202510090539.9A CN202510090539A CN119632550B CN 119632550 B CN119632550 B CN 119632550B CN 202510090539 A CN202510090539 A CN 202510090539A CN 119632550 B CN119632550 B CN 119632550B
- Authority
- CN
- China
- Prior art keywords
- data
- module
- motion
- arm
- acceleration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
本发明公开了一种基于单一加速度计的多功能系统，用于可穿戴设备上，包括：加速度计模块：采集用户手臂的三轴加速度数据；数据采集与预处理模块：对采集到的三轴加速度数据进行向量范数计算、数据标准化处理，并根据不同功能提取相关的特征；阈值检验模块：根据手臂运动的加速度变化和角度特征进行阈值判断，过滤掉无效动作；LDA分类模块：通过线性判别分析模型对经过预处理和阈值检验的数据进行分类，识别不同的手臂动作；功能模块：包括抬腕亮屏与息屏、计步、运动模式识别和睡眠评估功能，根据分类结果进行对应的响应控制。通过单一加速度计，精准地实现了抬腕亮屏与息屏、计步、运动模式识别和睡眠评估等多项功能。
The present invention discloses a multifunctional system based on a single accelerometer for use on wearable devices, comprising: an accelerometer module for collecting three-axis acceleration data of the user's arm; a data acquisition and preprocessing module for performing vector norm calculation and data standardization on the collected three-axis acceleration data, and extracting relevant features according to different functions; a threshold test module for performing threshold judgment based on the acceleration change and angle characteristics of the arm movement, filtering out invalid movements; an LDA classification module for classifying the preprocessed and threshold-tested data through a linear discriminant analysis model to identify different arm movements; a functional module for including functions such as wrist-raising to turn on and off the screen, step counting, motion pattern recognition, and sleep assessment, and performing corresponding response control according to the classification results. Through a single accelerometer, multiple functions such as wrist-raising to turn on and off the screen, step counting, motion pattern recognition, and sleep assessment are accurately realized.
Description
Technical Field
The invention belongs to the technical field of wearable equipment, and relates to a multifunctional system based on a single accelerometer, in particular to a system for step counting, real-time wrist lifting, screen lighting, screen extinguishing, movement pattern recognition and sleep evaluation.
Background
With the rapid development of technology, wearable device technology has gradually entered a more efficient and intelligent stage, wherein smart watches and bracelets are typical representatives, and by virtue of portability and versatility, the wearable device technology has been widely used in daily life. These devices typically rely on multi-sensor systems, such as accelerometers, gyroscopes, and magnetometers, to perform a variety of functions including gesture recognition, movement pattern analysis, and health monitoring.
However, although the multi-sensor cooperation can significantly improve the accuracy and the comprehensiveness of data, the technical scheme also brings about a plurality of challenges. First, the complexity of the multi-sensor system increases the difficulty of hardware design and integration, leading to increased costs in development and production of the device. Meanwhile, the simultaneous operation of a plurality of sensors also obviously improves the power consumption of the equipment, which forms a serious test on the cruising ability and the user experience of the equipment.
In terms of function implementation, the traditional threshold method is widely applied to the implementation of interactive functions such as wrist lifting and screen brightening. However, this method is prone to false triggering in complex actions or environments, resulting in poor user experience. In addition, the existing step counting algorithm and motion pattern recognition technology also face the problem of insufficient precision. Step counting algorithms often rely on simple acceleration changes or gait patterns, and step counting is performed through fixed threshold or maximum detection based on vibration amplitude, so that the step counting algorithms are difficult to adapt to variable motion states (such as jogging and fast running), and particularly step counting mistakes or missing counting easily occur under low-frequency actions. Motion pattern recognition techniques typically rely on simple feature analysis and fixed rule classification, such as by analyzing the mean and fluctuation amplitude of acceleration data to distinguish between walking, running, stationary, etc. states. However, such rule classification methods exhibit large limitations in the face of diversified user behaviors, particularly in scenes where a plurality of motion patterns alternate, and are prone to confusion.
Therefore, the existing wearable device technology has obvious defects in the aspects of complexity of a multi-sensor system, limitation of a threshold method, accuracy of a step counting algorithm, accuracy of motion pattern recognition and the like. These challenges limit the performance and user experience of the device.
Disclosure of Invention
The present invention is directed to a multifunctional system based on a single accelerometer, which solves the problems set forth in the background art.
The invention aims at realizing the following technical scheme:
A single accelerometer-based multi-functional system for use on a wearable device, comprising:
The accelerometer module is used for collecting triaxial acceleration data of the arm of the user;
The data acquisition and preprocessing module is used for carrying out vector norm calculation and data standardization processing on the acquired triaxial acceleration data and extracting relevant characteristics according to different functions;
the threshold value checking module is used for carrying out threshold value judgment according to the acceleration change and the angle characteristic of the arm movement and filtering out invalid actions;
The LDA classifying module classifies the data subjected to preprocessing and threshold value inspection through a linear discriminant analysis model and identifies different arm actions;
the functional module comprises wrist lifting and screen lighting and screen resting, step counting, movement pattern recognition and sleep evaluation functions, and corresponding response control is carried out according to the classification result.
As a further improvement of an embodiment of the present invention, the data acquisition and preprocessing module includes a data processing module, where the data processing module acquires acceleration data in three-axis directions, calculates vector norms of the three-axis acceleration in each time window, normalizes the acquired data, ensures that data in different dimensions can be compared in the same range, and then performs feature extraction on the data normalized by the data in each time window.
As a further improvement of an embodiment of the invention, the threshold value checking module comprises a data window for time threshold value checking, the data after feature extraction is filtered by the data window to obtain effective data, and the acceleration change of the Z axis and the inclination angle change of the Y axis are detected aiming at the effective data to realize preliminary judgment on the lifting and dropping actions of the arm.
As a further improvement of an embodiment of the present invention, the LDA classification module performs classification training through a linear discriminant analysis model based on multidimensional features in acceleration data according to feature vectors of different actions, and generates a classification model for predicting arm actions of a user in real time.
As a further improvement of an embodiment of the present invention, the function module judges whether the current arm motion is wrist lifting or wrist hanging by using an LDA model to judge a screen-on or screen-off control, and specifically includes the following steps:
Step 1.1, a triaxial accelerometer with the frequency of 100Hz is adopted in the intelligent watch;
Step 1.2, acquiring triaxial acceleration data of an arm 100 times per second by an accelerometer, transmitting the triaxial acceleration data to a data processing module in real time to calculate a vector norm to obtain the vector norm of the acceleration, and obtaining a normalized characteristic value by a normalization processing formula, wherein the normalization processing formula is as follows:
Where feature i is the i-th original feature, MEAN i is the MEAN of feature i, SCALE i is the standard deviation of feature i, scaled_feature i is the normalized feature value;
step 1.3, extracting the angle and acceleration characteristics of the arm in each time window, and judging that the wrist lifting action is performed when the Z-axis acceleration is negative and the Y-axis inclination angle is smaller than 0.6;
And step 1.4, inputting the features extracted in the step 1.3 into an LDA classification module, and judging whether the wrist lifting and screen-lighting action or the wrist hanging and screen-extinguishing action is performed according to the LDA classification result.
As a further improvement of an embodiment of the present invention, the function module identifies whether the gait motion is current through the LDA model, and specifically includes the following steps:
step 2.1, the intelligent watch adopts a triaxial accelerometer with the frequency of 100 Hz;
step 2.2, acquiring triaxial acceleration data, transmitting the triaxial acceleration data to a data processing module in real time, analyzing acceleration signals through fast Fourier transform, and extracting frequency characteristics;
Step 2.3, inputting the frequency characteristics extracted in the step 2.2 into an LDA classification module, and identifying whether gait motion is present or not;
and 2.4, after the gait action is identified, identifying the peak value of each step through a local maximum value detection algorithm, removing outliers, performing frequency domain analysis to perform anti-interference treatment, and distinguishing running and walking by adopting moving average filtering.
As a further improvement of an embodiment of the present invention, the step of determining the motion pattern classification by the functional module is as follows:
Step 3.1, the intelligent watch adopts a triaxial accelerometer with the frequency of 100 Hz;
step 3.2, acquiring triaxial acceleration data once every three seconds, and extracting features of the data in each time window;
And 3.3, carrying out normalization processing on the features extracted in the step 3.2, wherein the formula is as follows:
wherein y represents the extracted feature, the unified processing of the feature value in different dimensions is ensured through normalization processing, and then the extracted feature is input into an LDA model to optimize the classification precision, wherein the formula is as follows:
wherein W is a coefficient matrix, X is a motion feature value, and L is an intercept vector;
And 3.4, judging each element of each output value, wherein more than 0 is regarded as positive correlation, less than 0 is regarded as negative correlation, screening is carried out by combining with a judgment rule of a motion mode, and the current motion state is judged based on continuous 20 times of statistical output to realize mode switching and response.
As a further development of an embodiment of the invention, the motion pattern is determined by the rule that when more than 2 positive correlations or all negative correlations in the output are less than a set threshold, the output is excluded from recognition, but the statistics are still accumulated.
As a further improvement of an embodiment of the present invention, the sleep evaluation function performed by the function module includes the following steps:
step 4.1, the intelligent watch monitors night movement of a user by adopting an accelerometer, and the sampling frequency of a sensor is set to be 50Hz;
Step 4.2, acquiring triaxial acceleration data of the arm of the user every 30 seconds, and monitoring the movement and movement frequency of the arm;
And 4.3, calculating the average square wavelength and standard deviation of the low-frequency motion characteristics extracted in the step 3.2, and further judging whether the user is in a deep sleep, shallow sleep or wake state by analyzing the motion amplitude, frequency and duration, wherein the formula of the average square wavelength is as follows:
the standard deviation formula is as follows:
wherein y k represents acceleration data, and n is the total number of data in the sliding window;
And 4.4, recording and analyzing according to the data obtained in the step 4.3 to generate a sleep report.
As a further improvement of an embodiment of the present invention, the threshold judgment criteria in the step 4.3 is as follows, according to the extracted low-frequency motion characteristics and the set threshold, wherein squarewavelength and s are combined, different thresholds are set, a score of the current motion intensity is obtained, and according to the average value of the score, whether the sleep state is shallow, deep or awake state is judged.
The wrist lifting and screen lighting device has the advantages that the wrist lifting and screen lighting device accurately achieves multiple functions of wrist lifting and screen lighting, step counting, movement pattern recognition, sleep evaluation and the like through a single accelerometer and combining with advanced signal processing, linear Discriminant Analysis (LDA) models and other existing machine learning technologies. The method not only simplifies the hardware structure and obviously reduces the power consumption and the system complexity, but also ensures that the equipment can still keep high-efficiency and accurate action recognition and state judgment under limited sensor input through an optimization algorithm.
Compared with the prior art relying on multiple sensors, the patent has the outstanding advantages of high-efficiency processing and classifying capabilities of limited input data, and more reliable motion recognition and state monitoring under complex environments and different user behavior modes are ensured through feature extraction and machine learning models. The technology not only improves the user experience of the equipment, but also enables the equipment to be more suitable for portable and low-power consumption application scenes, and greatly promotes the application of the single-sensor equipment in the intelligent wearable field.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It will be apparent to those skilled in the art from this disclosure that the drawings described below are merely exemplary and that other embodiments may be derived from the drawings provided without undue effort.
The structures, proportions, sizes, etc. shown in the present specification are shown only for the purposes of illustration and description, and are not intended to limit the scope of the invention, which is defined by the claims, so that any structural modifications, changes in proportions, or adjustments of sizes, which do not affect the efficacy or the achievement of the present invention, should fall within the ambit of the technical disclosure.
Fig. 1 is a schematic diagram of a system architecture according to the present invention.
Detailed Description
It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. The application will be described in detail below with reference to the drawings in connection with embodiments.
It is noted that all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs unless otherwise indicated.
In the present invention, unless otherwise indicated, the use of orientation terms such as "upper, lower, top, bottom" are generally with respect to the orientation shown in the drawings or with respect to the component itself in the vertical, vertical or gravitational direction, and likewise, for ease of understanding and description, "inner, outer" refer to inner, outer relative to the profile of the component itself, but such orientation terms are not intended to limit the invention.
A multifunctional system based on a single accelerometer realizes the functions of wrist lifting, screen lighting, screen resting, step counting, movement pattern recognition, sleep evaluation and the like through the single accelerometer. The scheme is designed aiming at wearable equipment, has the advantages of simple structure, low power consumption and high precision, and is suitable for long-time daily use. The technical scheme and the implementation steps of the invention will be described in detail below.
The system architecture of the present invention is shown in fig. 1, and mainly comprises the following modules:
accelerometer module 10 collects three-axis acceleration data of the user's arm as a primary input source.
The data acquisition and preprocessing module 20 performs vector norm calculation and data standardization processing on the acquired triaxial acceleration data, and extracts relevant features according to different functions.
And the threshold value checking module 30 is used for carrying out threshold value judgment according to the acceleration change and the angle characteristic of the arm movement and filtering out invalid actions.
The LDA classification module 40 classifies the preprocessed and threshold checked data through a linear discriminant analysis model to identify different arm actions, such as wrist lifting, wrist hanging, gait and movement modes.
The functional module 50 comprises wrist lifting and screen lighting and screen resting, step counting, movement pattern recognition and sleep evaluation functions, and performs corresponding response control according to the classification result.
The description of the data acquisition and preprocessing module 20 in this embodiment is as follows:
The accelerometer is arranged on the wristwatch device and is responsible for collecting acceleration data in three axial directions. Samples 100 times per second and transmits the data to an upper processing module (data processing module).
Vector norm calculation, namely, in each time window, the system calculates the vector norm of the triaxial acceleration, and the vector norm is used for reflecting the overall motion intensity of the arm.
Data normalization, namely normalizing acquired data to ensure that data with different dimensions can be compared in the same range.
And extracting 9 features including mean value, variance, root mean square value, peak value, direction change, spectrum energy, wavelength, waveform coefficient and average absolute value of acceleration in each time window for subsequent LDA classification.
The description of the threshold checking module 30 in this embodiment is as follows:
and judging the angle and the acceleration threshold, namely primarily judging the lifting and lowering actions of the arm by the system by detecting the acceleration change of the Z axis and the inclination angle change of the Y axis. For example, when the Z-axis acceleration is less than 0 and the Y-axis inclination absolute value is less than 0.6, the wrist lifting operation is determined.
Time threshold checking-to prevent frequent bright and off-screen, the system sets a time threshold with at least 15 data windows (about 1.5 seconds) between each action, and only with continuous valid data will the corresponding operation be triggered.
The description of the LDA classification module 40 in this embodiment is as follows:
And (3) training the LDA model, namely carrying out classification training by the system through a Linear Discriminant Analysis (LDA) model according to multidimensional features in the acceleration data, and generating a classification model for predicting the arm action of the user in real time. The training data of the model covers different arm actions, such as wrist lifting, wrist hanging, gait and the like.
And (3) classifying the data in each time window by the LDA model in real time during the running process of the system, and classifying the data into corresponding action categories such as wrist lifting, gait or stillness according to the characteristic vectors of different actions.
The linear discriminant analysis (English: LINEARDISCRIMINANTANALYSIS, abbreviated: LDA) in this example is a widely used data analysis method, and is commonly used for feature degradation and classification.
The description of the functional module 50 in this embodiment is as follows:
1. Wrist lifting bright screen and screen extinguishing function
And (3) screen-lighting control, namely when the LDA model judges that the current arm motion is the wrist lifting motion, the system confirms the effectiveness of the motion through threshold value inspection and triggers the screen to be lightened.
And (3) screen-off control, namely when the wrist hanging action (LDA classification result is wrist hanging type) is detected and the threshold value is checked, the system is used for triggering the screen to be closed. And through setting of the time threshold, frequent screen-lighting and screen-extinguishing operations are prevented, and electric power is saved.
2. Step counting function
Gait signal detection the system identifies gait signals by frequency domain analysis of the acceleration signals. FFT (fast fourier transform) is used to extract gait features such as step frequency and amplitude.
LDA gait classification the system inputs gait data into the LDA model and classifies the motion as gait or other motion, such as running, jogging, etc., by feature analysis.
And step number calculation, namely combining a local maximum detection algorithm, and calculating the step number in real time by analyzing the step change when the system recognizes gait. For non-gait signals (such as rest or disturbance actions), the system filters them out, avoiding false counts.
3. Motion pattern recognition function
Motion feature extraction, namely extracting features under different motion modes, such as walking, running, stillness and the like, by the system through acceleration data. The extracted features include acceleration mean, standard deviation, direction change, etc.
And the LDA classifier classifies the extracted features by using an LDA model, and the system can accurately identify different motion modes and perform corresponding processing according to classification results, such as recording motion data or prompting the current motion state of a user.
4. Sleep assessment function
And (3) analyzing the low-frequency motion characteristics, namely, at night, judging the sleep state of the user by analyzing the low-frequency motion characteristics of the arm by the system. Sleep states are classified as deep sleep, shallow sleep, and wake, and the system is able to distinguish between these states based on the stationarity and frequency variation of the acceleration data.
Time threshold in order to avoid misjudgment in a short time, the system analyzes the states of a plurality of continuous data windows, ensures the stability of state change and generates a detailed sleep report.
In summary, the system architecture of the present invention has the following functions:
And the data acquisition, namely the accelerometer module acquires triaxial acceleration data in real time.
And data preprocessing, namely carrying out vector norm calculation and standardization processing on the data by the system, and extracting features.
Threshold judgment and LDA classification, namely, the system identifies the action of the arm of the user through an LDA model and threshold inspection.
The function is realized by respectively triggering wrist lifting, screen brightening, step counting, movement pattern recognition or sleep evaluation functions by the system according to the classification result.
According to the technical scheme, the intelligent watch system integrating the functions of step counting, wrist lifting, screen lightening, screen extinguishing, movement pattern recognition and sleep evaluation is provided through a single accelerometer and a signal processing and linear discriminant analysis model (LDA model).
Compared with the prior art, the invention has obvious technical progress in a plurality of aspects, and brings the following beneficial effects:
1. system complexity and power consumption are reduced, and equipment endurance is improved
The prior art generally relies on a multi-sensor system, increases hardware complexity and energy consumption, and affects portability and user experience of the device. The invention realizes the multifunctional application by using only a single accelerometer and combining with an optimized signal processing technology. Because a plurality of sensors are not required to work simultaneously, the power consumption of the system is obviously reduced, the equipment can be kept to operate efficiently for a long time, and the cruising experience of a user is improved. In addition, the hardware structure of the system is simpler, the design and manufacturing cost is reduced, and the equipment has obvious advantages in portability and performance.
2. Accuracy of wrist lifting bright screen and screen extinguishing function is improved, and false triggering is reduced
The existing wrist lifting and screen lighting and screen extinguishing functions depend on a simple threshold method, and are easy to trigger by mistake in a complex action scene, so that the screen is frequently lighted and extinguished. According to the invention, by combining the LDA model and the multidimensional feature extraction, the effective wrist lifting action can be accurately identified, and meanwhile, the false triggering condition is remarkably reduced through time and angle threshold value inspection. Because the system can adapt to the action change of different users, the users can not frequently consume batteries due to false triggering in daily use, and the efficiency and the user experience of the equipment are further improved.
3. Remarkably improves the precision of the step counting function
The step counting algorithm in the prior art is easy to appear the phenomenon of step counting missing or false counting under different gait (such as slow walking, fast running and the like), and particularly has larger step counting error under a complex environment. According to the invention, gait signals are accurately identified through LDA model classification, and the high-precision step number calculation can be kept under a changeable motion state by combining a local maximum detection algorithm. Whether slow walking, fast walking or running is performed, the method can ensure the accuracy of step number recording through the feature extraction and classification model, avoid missing counting and false counting, and meet the high requirement of users on motion data.
4. Improving reliability and adaptability of motion pattern recognition
Most of the existing motion pattern recognition systems depend on rule classification algorithms, cannot effectively cope with complex motion scenes, and are easy to confuse different motion patterns. The invention can accurately distinguish different movement modes such as walking, running, stillness and the like through the multidimensional feature analysis based on the LDA model. Because the LDA model can automatically learn the characteristics of the motion mode from the data, the system can adapt to diversified user behaviors, errors and confusion in mode recognition are reduced, and the overall recognition accuracy and stability of the system are improved.
5. Optimizing accuracy and power consumption of sleep assessment functions
Existing sleep assessment systems typically rely on multiple sensors or complex machine learning models, and although with high accuracy, power consumption is also relatively large, especially in long term use, battery consumption of the device becomes a major issue. According to the invention, by analyzing the low-frequency motion characteristics of the accelerometer, the sleeping state of the user, such as deep sleep, shallow sleep and awakening, can be judged without depending on a complex machine learning model. The method reduces the power consumption of the system, ensures the accuracy of sleep evaluation, and enables the equipment to provide a reliable sleep monitoring report under the condition of low power consumption.
6. Improving the cost effectiveness of the device
Because the invention realizes multiple functions through the single accelerometer, the dependence on other sensors is reduced, the system hardware structure is simplified, and the design and manufacturing cost is obviously reduced. The overall cost effectiveness of the equipment is greatly improved, and the equipment has stronger market competitiveness. In addition, the simplified design also shortens the production period of the equipment, and further improves the economic benefit.
In summary, the invention greatly improves the accuracy of functions and the user experience while reducing the complexity of the system and the power consumption by optimizing the hardware structure and the algorithm design. The technical scheme not only has remarkable advantages in the aspects of wrist lifting, screen brightening, step counting, movement pattern recognition and sleep evaluation, but also provides a new direction for the development of intelligent wearable equipment in the future.
The invention provides various intelligent wristwatch realization modes based on a single accelerometer, and the functions of step counting, movement pattern recognition, wrist lifting, screen brightening, screen-off sleep monitoring and the like are realized through the organic cooperation of a plurality of algorithms. The method not only designs an algorithm with strong adaptability in each functional module, but also combines the technologies of anti-interference processing, feature extraction, LDA classification, statistical analysis and the like, so that the system has higher reliability and precision in a complex environment. The following examples detail the specific implementation steps and conditions of these functional modules.
Example 1 wrist lifting and Screen brightening and Screen extinguishing function
Step 1.1 hardware preparation
Device-use of a smart watch with a three-axis accelerometer.
The accelerometer is embedded in the wristwatch, the sampling frequency of the sensor is set to be 100Hz, and high-precision detection of arm movement is ensured.
Step 1.2 data acquisition and processing
Data acquisition, namely acquiring three-axis acceleration data of the arm 100 times per second by using an accelerometer, and transmitting the three-axis acceleration data to a data processing module in real time.
Vector norm calculation the vector norm of the acceleration is calculated by the following formula:
where v is a vector comprising a plurality of components a x,ay,az, each of which is squared and summed, and the result is square-root-manipulated to obtain the euclidean norm. The vector norm is used to measure the magnitude of the overall motion of the arm.
The normalization processing formula is used for scaling the data to obtain a MEAN value array and a standard deviation SCALE array which take the feature quantity as a SCALE through training set training on the same SCALE, wherein the formula is as follows:
Where feature i is the i-th original feature, MEAN i is the MEAN of feature i, SCALE i is the standard deviation of feature i (which may also be the range of features or other scaling value), and scaled_feature i is the normalized feature value.
Step 1.3, feature extraction and threshold judgment
Extracting features, namely extracting the angle and acceleration features of the arm in each time window, and focusing on the change of the Z-axis acceleration (consistent with the vertical direction).
And judging the threshold value, namely preliminarily judging that the wrist lifting action is performed when the Z-axis acceleration is negative and the Y-axis inclination angle is smaller than 0.6. By setting a time threshold (e.g., at least 1.5 seconds per action interval), frequent triggering of a lighted screen is avoided.
Step 1.4 LDA Classification and control
LDA classification, namely inputting the extracted features into an LDA classification module, and identifying whether the wrist lifting or wrist hanging actions are performed by a classifier.
And (3) controlling the screen to be lightened and the screen to be extinguished, namely triggering the screen to be lightened after the system confirms that the wrist lifting is effective according to the LDA classification result, and triggering the screen extinguishing function if the classification result is wrist hanging.
Example 2 step-counting function
Step 2.1 hardware preparation
Device-also using a smart watch with an accelerometer, the sampling frequency of the sensor is kept at 100Hz.
Step 2.2 data acquisition and Signal processing
Data acquisition, namely, the watch acquires triaxial acceleration data and transmits the triaxial acceleration data to the data processing module. The acceleration signal is analyzed by FFT (fast fourier transform), and frequency characteristics are extracted for gait recognition.
Step 2.3 feature extraction and gait recognition
Feature extraction, namely extracting features such as frequency, amplitude, stride and the like in gait signals as a basis for gait recognition.
LDA classification-inputting the extracted features into an LDA classification module, identifying whether or not a gait action, such as walking or running, is currently performed.
Step 2.4 step count calculation
Local maximum detection after the gait action is identified, the system identifies the peak of each step by a local maximum detection algorithm. The system updates the step count in real time based on the detected step count increment.
And anti-interference processing, namely removing outliers, then carrying out frequency domain analysis, and screening a few irregular interference actions through a threshold value. The specific interference action is filtered through triaxial acceleration feature extraction, running and walking are distinguished through moving average filtering, and gait recognition accuracy is ensured.
Example 3 motion Pattern recognition functionality
Step 3.1 hardware preparation
Device-using a smart watch with an accelerometer, the sampling frequency of the sensor is set to 100Hz.
Step 3.2 data acquisition and processing
Data acquisition, namely continuously acquiring triaxial acceleration data by the watch, and extracting characteristics of the data in each time window;
And 3, motion pattern recognition, namely extracting triaxial acceleration signals at one time for 3 s.
Step 3.3 motion Pattern Classification
Feature extraction, namely extracting multidimensional features such as acceleration mean value, direction change, standard deviation and the like of the arm, and carrying out normalization processing without filter processing:
where y represents the various features extracted, which ensures that feature values are uniformly processed within different dimensions.
LDA classification, input the extracted features into LDA model, through the formula:
wherein W is a coefficient matrix, X is a motion characteristic value, and L is an intercept vector, so that classification accuracy is further optimized. Unlike conventional probability transitions, each output value is directly judged, and rule-based classification is performed.
Step 3.4, positive and negative correlation determination and mode switching
Adaptive determination of positive and negative correlation by determining for each element of each output value, more than 0 is considered positive and less than 0 is negative, and screening is performed in combination with a specific decision rule of the motion pattern. For example, when more than 2 positive correlations or all negative correlations in the output are less than a set threshold, the output is excluded from recognition, but the statistics are still accumulated.
Mode switch and response the system determines the current state of motion based on 20 consecutive statistical outputs. The judgment mode based on the accumulated statistics enhances the stability of the recognition, reduces erroneous judgment in a short period and improves the accuracy of the system in long-time motion recognition. The system will record the changes and perform corresponding data processing, such as athletic data recording and prompting the user for the current athletic status.
Embodiment 4 sleep evaluation function
Step 4.1 hardware preparation
Device-use of an accelerometer to monitor the night movement of the user, the sampling frequency of the sensor being set to 50Hz.
Step 4.2 data acquisition and processing
Data acquisition during the night, the watch continuously acquires triaxial acceleration data of the arm of the user, and mainly monitors the micro-motion and the motion frequency of the arm. Sleep evaluation, namely extracting triaxial acceleration signals once for 30 seconds.
Step 4.3, feature extraction and sleep State determination
Feature extraction, namely extracting low-frequency motion features of the arm, including average square wavelength:
Standard deviation:
where y k represents acceleration data, and n is the total number of data in the sliding window. By analyzing the amplitude, frequency and duration of the motion, it is determined whether the user is in a deep sleep, shallow sleep or awake state.
Threshold determination based on the extracted low frequency motion features and set thresholds, the system determines the sleep stage of the user. And integrating squarewavelength and s, and setting different thresholds to obtain a score of the current exercise intensity. The sleep mode comprises the steps of judging to enter sleep if the average value of the score for a long period is lower than a threshold value 1, judging to enter the sleep to be light sleep by default, judging to enter deep sleep if the average value of the score for a period is lower than a threshold value 2 after the light sleep is over a specified period, judging to enter the light sleep if the average value of the score for a period is higher than a threshold value 3 after the deep sleep is over a period, and judging to enter the wake if the average value of the score for a period is higher than a threshold value 4 in a light sleep or deep sleep state.
Step 4.4 sleep report Generation
Data logging and analysis the system logs the sleep state of the user overnight and generates detailed sleep reports including deep sleep, shallow sleep and wake time. The report may be used to analyze the sleep quality of the user and provide personalized health advice.
Algorithm matching and technical effect summary
The four embodiments not only realize functions through respective independent algorithms, but also realize technical advantages of the invention through cooperative work among algorithms:
1. The throughout application of the anti-interference processing is that whether the wrist is lifted, the screen is lightened, the step is counted or the motion mode is identified, the anti-interference processing (such as outlier removal, frequency domain analysis and threshold judgment) is embodied in all functional modules, and the stability under the complex environment is ensured.
The unified application of the LDA classification module is that the LDA model is widely applied to wrist lifting bright screen, gait recognition and motion pattern recognition. Through unified feature extraction and classification strategies, the architecture of the system is simplified, and the recognition accuracy and response speed are improved.
3. The multi-level algorithm effectively cooperates with a plurality of algorithms such as feature extraction, frequency domain analysis, moving average filtering, local maximum detection and the like, so that the high efficiency of the system in data acquisition, processing, identification and control is ensured.
4. The statistics-based erroneous judgment correction mechanism effectively reduces short-term erroneous judgment by a judgment mode (such as 20 times of statistics rules) of multiple statistics and accumulated output, and ensures the stability of the system in long-time use.
Through the organic cooperation of the algorithms, the intelligent wrist watch multifunctional device not only improves the multifunctional realization capability of the intelligent wrist watch, but also reduces the power consumption and the hardware complexity through a unified algorithm architecture, and provides an ideal technical scheme for multifunctional and low-power wearable equipment.
The technical scheme of the invention realizes a plurality of functions through a single accelerometer, including step counting, wrist lifting, screen lighting, screen extinguishing, movement pattern recognition and sleep evaluation, and has the characteristics of hardware design simplification, low power consumption and high precision. Through the organic coordination of a plurality of algorithms, such as anti-interference processing, LDA classification, feature extraction and time threshold management, the system can stably operate in a complex environment. The following key technical points and the technical scheme needing key protection of the invention are as follows:
Technical scheme for realizing multiple functions based on single accelerometer
The key points of the technology are as follows:
1. three-axis acceleration data are acquired by using a single accelerometer, and multiple functional support (wrist lifting, step counting, movement pattern recognition and sleep evaluation) is provided through feature extraction and signal processing.
2. Features are extracted from the acceleration data by signal processing techniques such as vector norm calculations, FFT frequency domain analysis, etc.
3. And combining a Linear Discriminant Analysis (LDA) model to classify and identify the actions of the multifunctional sensor, so as to realize a high-precision function based on limited sensor input.
4. The anti-interference algorithm (outlier removal and frequency domain analysis) and the feature extraction algorithm are applicable to all functional modules, so that the influence of noise and interference on the system performance is reduced.
Precise identification and control method for wrist lifting bright screen and idle screen
The key points of the technology are as follows:
1. And the wrist lifting and wrist hanging actions are primarily identified by extracting the characteristics of the Z-axis acceleration and the Y-axis inclination angle and combining threshold judgment.
2. The wrist lifting action is accurately classified by using the LDA model, so that the accuracy of action identification is ensured, and the false triggering phenomenon is reduced.
3. Setting a time threshold value, and avoiding frequent switching between bright screen and off screen.
4. Erroneous judgment is reduced by continuously counting the classification result for a plurality of times (such as 20 times of accumulated statistics), and system stability is enhanced.
Third, high-precision step counting algorithm based on LDA model
The key points of the technology are as follows:
1. gait signals are acquired by an accelerometer and gait features (e.g., step frequency, amplitude, etc.) are extracted using FFT frequency domain analysis.
2. And classifying gait signals by using an LDA model, calculating the number of steps by combining a local maximum detection algorithm, and reducing the influence of interference signals on the step number calculation.
3. Real-time gait classification and error correction ensure the step counting precision under various gaits (slow walking, fast walking and running).
4. The moving average filter distinguishes walking from running, and the accuracy of step number calculation is improved by combining a local maximum detection algorithm.
Motion pattern recognition method
The key points of the technology are as follows:
1. multidimensional features (such as mean value, standard deviation, direction change and the like) in the accelerometer are extracted, and the movement mode classification is carried out by combining with the LDA model to distinguish movement states of walking, running, stillness and the like.
2. The method is dynamically adapted to the movement behavior modes of different users, and ensures that different movement states are accurately identified in a complex movement scene.
3. And each element output is directly judged without using a traditional probability conversion step, and positive correlation and negative correlation are classified by combining a specific rule. Compared with probability conversion, the method reduces the computational complexity and improves the real-time response speed and the recognition accuracy.
4. Based on the motion classification and positive-negative correlation judgment of the LDA model, false judgment in a short period is reduced through continuous multiple statistical output, and the recognition stability is improved.
Fifth, sleep evaluation function with low power consumption
The key points of the technology are as follows:
1. and extracting the micro motion characteristics of the arms through the low-frequency acceleration signals, and judging the sleep state (deep sleep, shallow sleep and awakening) of the user.
2. And the method does not depend on a complex machine learning algorithm, accurately judges different sleep stages by setting a threshold value and a time window, and generates a sleep report.
3. The low-power-consumption algorithm ensures long-time continuous monitoring and realizes efficient sleep evaluation and health advice generation.
Sixth time threshold and false triggering prevention mechanism
The key points of the technology are as follows:
1. By setting the time threshold, frequent triggering is avoided in each function (such as wrist lifting, screen lighting, screen extinguishing and step counting) and the stability of the system is ensured.
2. And combining action classification and time window management, preventing the system from excessively responding to short-time or invalid actions of a user, and reducing misoperation.
3. The time threshold value is combined with action classification, so that the system response speed and stability are improved while false triggering is reduced.
Seventh, integration and implementation of the overall system
The key points of the technology are as follows:
1. The whole system integrates a single accelerometer, a multifunctional signal processing module, an LDA classification model and control logic to form the multifunctional wearable device with a simplified hardware structure.
2. The system architecture design combines the cooperative work of a plurality of functions, reduces the dependence of multiple sensors and realizes the balance of functions and performances.
3. By sharing accelerometer data, the data among different functional modules are efficiently utilized and flexibly switched, and the overall performance of the system is improved.
Conclusion(s)
The technical key points and the protection points of the invention mainly focus on using a single accelerometer, and realize multifunctional application such as step counting, wrist lifting and screen lighting and screen extinguishing, movement pattern recognition and sleep evaluation through signal processing and an LDA model. The intelligent wristwatch system with high precision, low power consumption and portability is provided by simplifying the hardware structure, reducing the power consumption, introducing anti-interference processing and feature extraction strategies and combining multidimensional feature extraction and a machine learning classification model.
It will be apparent that the embodiments described above are merely some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present application. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A single accelerometer-based multi-functional system for use on a wearable device, comprising:
The accelerometer module is used for collecting triaxial acceleration data of the arm of the user;
The data acquisition and preprocessing module is used for carrying out vector norm calculation and data standardization processing on the acquired triaxial acceleration data and extracting relevant characteristics according to different functions;
the threshold value checking module is used for carrying out threshold value judgment according to the acceleration change and the angle characteristic of the arm movement and filtering out invalid actions;
The LDA classifying module classifies the data subjected to preprocessing and threshold value inspection through a linear discriminant analysis model and identifies different arm actions;
The functional module comprises wrist lifting and screen lighting and screen resting, step counting, movement pattern recognition and sleep evaluation functions, and performs corresponding response control according to the classification result;
The function module identifies whether the current gait action is the gait action through the LDA model, and specifically comprises the following steps:
step 2.1, the intelligent watch adopts a triaxial accelerometer with the frequency of 100 Hz;
step 2.2, acquiring triaxial acceleration data, transmitting the triaxial acceleration data to a data processing module in real time, analyzing acceleration signals through fast Fourier transform, and extracting frequency characteristics;
Step 2.3, inputting the frequency characteristics extracted in the step 2.2 into an LDA classification module, and identifying whether gait motion is present or not;
step 2.4, after the gait action is identified, identifying the peak value of each step through a local maximum value detection algorithm, removing outliers, performing frequency domain analysis to perform anti-interference treatment, and distinguishing running and walking by adopting moving average filtering;
The function module judges the classification steps of the motion modes as follows:
Step 3.1, the intelligent watch adopts a triaxial accelerometer with the frequency of 100 Hz;
step 3.2, acquiring triaxial acceleration data once every three seconds, and extracting features of the data in each time window;
And 3.3, carrying out normalization processing on the features extracted in the step 3.2, wherein the formula is as follows: ;
wherein y represents the extracted feature, the unified processing of the feature value in different dimensions is ensured through normalization processing, and then the extracted feature is input into an LDA model to optimize the classification precision, wherein the formula is as follows:
;
Wherein the method comprises the steps of Is a matrix of coefficients that are selected from the group consisting of,The motion characteristic value is used for determining the motion characteristic value,Is an intercept vector;
step 3.4, judging each element of each output value, wherein more than 0 is regarded as positive correlation, and less than 0 is regarded as negative correlation, and screening by combining with a judgment rule of a motion mode;
The motion pattern is determined by excluding the output from recognition when more than 2 positive correlations or all negative correlations are less than a set threshold, but accumulating the statistics.
2. The system of claim 1, wherein the data acquisition and preprocessing module comprises a data processing module, the data processing module acquires acceleration data in three-axis directions, calculates vector norms of the three-axis acceleration in each time window, normalizes the acquired data, ensures that data in different dimensions can be compared in the same range, and performs feature extraction on the data normalized by the data in each time window.
3. The multifunctional system of claim 1, wherein the threshold inspection module comprises a data window for time threshold inspection, the data after feature extraction is filtered by the data window to obtain effective data, and the acceleration change of the Z axis and the inclination change of the Y axis are detected for the effective data to realize preliminary judgment on the lifting and lowering actions of the arm.
4. The system of claim 1, wherein the LDA classification module performs classification training by a linear discriminant analysis model based on multidimensional features in the acceleration data according to feature vectors of different actions to generate a classification model for predicting arm actions of a user in real time.
5. The multi-functional system of claim 1, wherein the function module determines whether the current arm motion is wrist lifting or wrist hanging by using an LDA model to determine a screen-on or screen-off control, and the method specifically comprises the following steps:
Step 1.1, a triaxial accelerometer with the frequency of 100Hz is adopted in the intelligent watch;
Step 1.2, acquiring triaxial acceleration data of an arm 100 times per second by an accelerometer, transmitting the triaxial acceleration data to a data processing module in real time to calculate a vector norm to obtain the vector norm of the acceleration, and obtaining a normalized characteristic value by a normalization processing formula, wherein the normalization processing formula is as follows: ;
Wherein the method comprises the steps of Is the firstThe number of original features is chosen to be,Is characterized byIs used for the average value of (a),Is characterized byIs set in the standard deviation of (2),Is a normalized characteristic value;
step 1.3, extracting the angle and acceleration characteristics of the arm in each time window, and judging that the wrist lifting action is performed when the Z-axis acceleration is negative and the Y-axis inclination angle is smaller than 0.6;
And step 1.4, inputting the features extracted in the step 1.3 into an LDA classification module, and judging whether the wrist lifting and screen-lighting action or the wrist hanging and screen-extinguishing action is performed according to the LDA classification result.
6. The multi-function system of claim 1, wherein the function module performs sleep assessment function steps of:
step 4.1, the intelligent watch monitors night movement of a user by adopting an accelerometer, and the sampling frequency of a sensor is set to be 50Hz;
Step 4.2, acquiring triaxial acceleration data of the arm of the user every 30 seconds, and monitoring the movement and movement frequency of the arm;
And 4.3, calculating the average square wavelength and standard deviation of the low-frequency motion characteristics extracted in the step 3.2, and further judging whether the user is in a deep sleep, shallow sleep or wake state by analyzing the motion amplitude, frequency and duration, wherein the formula of the average square wavelength is as follows: ;
the standard deviation formula is as follows: ;
Wherein, the Representing the data of the acceleration of the vehicle,Is the total number of data in the sliding window;
And 4.4, recording and analyzing according to the data obtained in the step 4.3 to generate a sleep report.
7. The multi-functional system of claim 6, wherein the threshold determination criteria in step 4.3 is as follows, wherein the threshold is set according to the extracted low frequency motion characteristics and the set threshold, wherein squarewavelength and s are combined, different thresholds are set, score for the current motion intensity is obtained, and the state of shallow sleep, deep sleep or wakefulness is determined according to the average value of score.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510090539.9A CN119632550B (en) | 2025-01-21 | 2025-01-21 | A multifunctional system based on a single accelerometer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510090539.9A CN119632550B (en) | 2025-01-21 | 2025-01-21 | A multifunctional system based on a single accelerometer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN119632550A CN119632550A (en) | 2025-03-18 |
| CN119632550B true CN119632550B (en) | 2025-10-24 |
Family
ID=94951484
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510090539.9A Active CN119632550B (en) | 2025-01-21 | 2025-01-21 | A multifunctional system based on a single accelerometer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119632550B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117037657A (en) * | 2023-06-28 | 2023-11-10 | 深圳小澈科技有限公司 | Display control method, intelligent watch and readable medium |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101789462B1 (en) * | 2012-01-19 | 2017-10-23 | 나이키 이노베이트 씨.브이. | Energy expenditure |
| WO2014144015A2 (en) * | 2013-03-15 | 2014-09-18 | Keller Eric Jeffrey | Computing interface system |
| WO2015021223A1 (en) * | 2013-08-07 | 2015-02-12 | Nike Innovate C.V. | Wrist-worn athletic device with gesture recognition and power management |
| US20160018899A1 (en) * | 2014-07-18 | 2016-01-21 | Apple Inc. | Detecting loss of user focus in a device |
| US9808185B2 (en) * | 2014-09-23 | 2017-11-07 | Fitbit, Inc. | Movement measure generation in a wearable electronic device |
| US20160089080A1 (en) * | 2014-09-30 | 2016-03-31 | Mophie, Inc. | System and method for activity determination |
| CN109938739B (en) * | 2017-12-20 | 2022-07-12 | 深圳先进技术研究院 | Cervical vertebra monitoring device |
| CN112130678B (en) * | 2020-09-23 | 2023-06-02 | 深圳市爱都科技有限公司 | Arm drop recognition method and wearable terminal |
| CN114298105B (en) * | 2021-12-29 | 2023-08-22 | 东莞市猎声电子科技有限公司 | Signal processing method for quickly responding to wrist lifting action and brightening screen in running process |
| CN116225227B (en) * | 2023-03-10 | 2024-07-09 | 东莞市猎声电子科技有限公司 | Method for judging bright screen based on finite state machine |
| CN117731238A (en) * | 2023-12-08 | 2024-03-22 | 成都天奥电子股份有限公司 | Sleep stage detection method, system and readable storage medium |
-
2025
- 2025-01-21 CN CN202510090539.9A patent/CN119632550B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117037657A (en) * | 2023-06-28 | 2023-11-10 | 深圳小澈科技有限公司 | Display control method, intelligent watch and readable medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN119632550A (en) | 2025-03-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102640085B (en) | For identifying the system and method for gesture | |
| CN105184325B (en) | Mobile intelligent terminal | |
| EP2802255B1 (en) | Activity classification in a multi-axis activity monitor device | |
| CN112464738B (en) | Improved naive Bayes algorithm user behavior identification method based on mobile phone sensor | |
| TW202036245A (en) | System for recognizing user gestures according to mechanomyogram detected from user’s wrist and method thereof | |
| CN106919958B (en) | A method of human finger motion recognition based on smart watch | |
| CN111178155B (en) | Gait feature extraction and gait recognition method based on inertial sensor | |
| CN106228200B (en) | Action identification method independent of action information acquisition equipment | |
| CN111089604B (en) | Fitness motion recognition method based on wearable sensor | |
| WO2018040757A1 (en) | Wearable device and method of using same to monitor motion state | |
| CN110807471B (en) | A multi-modal sensor behavior recognition system and recognition method | |
| CN108549900A (en) | Tumble detection method for human body based on mobile device wearing position | |
| CN108958482B (en) | Similarity action recognition device and method based on convolutional neural network | |
| CN111643092A (en) | Epilepsy alarm device and epilepsy detection method | |
| CN107085367A (en) | A kind of lift hand bright screen method of intelligent watch | |
| CN119632550B (en) | A multifunctional system based on a single accelerometer | |
| CN120738896B (en) | Abnormal detection methods for fabric processing equipment, electronic equipment and fabric processing equipment | |
| CN114298105B (en) | Signal processing method for quickly responding to wrist lifting action and brightening screen in running process | |
| CN109620241B (en) | Wearable device and motion monitoring method based on same | |
| CN113670432B (en) | Vibration information self-perception identification and self-energy supply sensing system identification method | |
| CN111666780A (en) | Intelligent door control security method based on emotion recognition technology | |
| Kareem et al. | Multimodal approach for fall detection based on support vector machine | |
| Vecchio et al. | Gait-based authentication: Evaluation of energy consumption on commercial devices | |
| WO2018014432A1 (en) | Voice application triggering control method, device and terminal | |
| CN120345861B (en) | Sleep staging method and system based on adaptive trigger time characteristics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |