Search

CN-122019964-A - Intelligent recording method for motion data

CN122019964ACN 122019964 ACN122019964 ACN 122019964ACN-122019964-A

Abstract

The invention provides an intelligent motion data recording method and a corresponding system 1, wherein the intelligent motion data recording method comprises the following steps of S1, performing time synchronization and denoising pretreatment on original motion data from a multi-source sensor, S2, dividing the pretreated multi-dimensional motion data into a plurality of overlapped sliding windows according to time sequence and caching the overlapped sliding windows in a memory, S3, sequentially executing multi-dimensional characteristic judgment on data points except for the head and the tail in each sliding window according to preset priority, marking the current point and terminating judgment of the other dimensions once any dimension judges the current point as a characteristic point, and realizing efficient, real-time and high-fidelity motion data thinning on resource-limited equipment through a sliding window structure, multi-dimensional priority judgment, a dynamic threshold mechanism and a forced endpoint reservation strategy in practical application, so that all technical indexes set by the purpose of the invention are completely met.

Inventors

  • NIU BENHUA
  • PENG JINAN

Assignees

  • 深圳市汇泰科电子有限公司

Dates

Publication Date
20260512
Application Date
20251225

Claims (10)

  1. 1. An intelligent recording method of motion data is characterized by comprising the following steps, S1, carrying out time synchronization and denoising pretreatment on original motion data from a multi-source sensor; S2, dividing the preprocessed multidimensional motion data into a plurality of overlapped sliding windows according to time sequence, and caching the sliding windows in a memory; s3, executing multidimensional feature judgment on the data points except the head and the tail in each sliding window according to preset priorities in sequence, and marking the current point and terminating judgment of other dimensions once any dimension judges the current point as a feature point; s4, forcedly reserving a first group of data points and a last group of data points of each sliding window as characteristic points; And S5, only writing all the data marked as the feature points into a local storage or uploading the data to a remote server, and discarding the rest of original data.
  2. 2. The intelligent exercise data recording method according to claim 1, wherein in the step S1, the time synchronization process interpolates and aligns heart rate, speed, pedal frequency, power and altitude data based on a time stamp output by the GPS module; The denoising treatment adopts first-order low-pass filtering, and the cut-off frequency is set to be 0.5Hz to 2Hz.
  3. 3. The intelligent recording method of motion data according to claim 2, wherein in the step S2, the sliding window is composed of n groups of multi-dimensional data after time alignment, and each group of data comprises six dimensions of values of longitude and latitude, heart rate, speed, pedaling frequency, power and altitude of a GPS at the same time; Overlapping of one data point exists between adjacent sliding windows; The value range of the sliding window length n is 5 to 30, and the sliding window length n is dynamically adjusted according to the current available memory capacity of the equipment, the CPU main frequency and the residual battery capacity.
  4. 4. The intelligent motion data recording method according to claim 3, wherein the sliding window length n is adjusted to 5 to 10 when the device is in a low power consumption mode or the memory margin is lower than a preset threshold value, and is extended to 15 to 30 when the device resources are sufficient.
  5. 5. The method of claim 1, wherein in step S3, the multi-dimensional feature determination is performed in order of priority, namely, altitude dimension, physiological and motion parameter dimension, GPS track dimension, wherein the physiological and motion parameter dimension comprises heart rate, speed, pedaling frequency and power.
  6. 6. The method of claim 5, wherein in determining the altitude dimension, an absolute value ΔH= |Hi-H i-1 | of an altitude difference between a current point and a previous point is calculated, and if ΔH > 2m, the i-th group of data is marked as a feature point.
  7. 7. The method of claim 5, wherein in the judgment of physiological and motion parameter dimensions, for any parameter Y ε { heart rate, velocity, pedaling frequency, power }, a current point P 1 (Xi, yi) and its front and rear adjacent points P 0 (X i-1 ,Y i-1 )、P 2 (X i+1 ,Y i+1 are taken to construct a straight line segment P 0 P 2 , and a vertical distance D from the point P 1 to the straight line is calculated; And if D exceeds the dynamic threshold T_Y, marking the ith group of data as a characteristic point, wherein the dynamic threshold T_Y is adaptively adjusted according to the current motion intensity.
  8. 8. The intelligent recording method of motion data according to claim 7, wherein when the speed V is greater than 20km/h or the power P is greater than 300W, T_Y=T 0 ×1.5, and when the speed V is less than or equal to 10km/h and the power P is less than or equal to 150W, T_Y=T 0 ×0.6; In the intermediate state, T_Y is linearly interpolated between 0.6T 0 and 1.5T 0 , wherein T 0 is a reference threshold value, the value range is 0.8 to 1.5, and the units are dependent on parameters.
  9. 9. The method of claim 5, wherein in the step of determining the GPS track dimension, P 0 is set as a last confirmed feature point, P 1 is set as a last confirmed feature point, P 2 is a current point to be determined, the true geographic vertical distance d_geo from P 2 to a line segment P 0 P 1 is calculated, if d_geo > T_gps, P 2 is marked as a new feature point, and the previous group of original data points are synchronously marked as inflection point features, and T_gps is adaptively adjusted according to the local track curvature kappa.
  10. 10. The intelligent motion data recording method according to claim 9, wherein the curvature kappa is calculated by three points P a-1 、P a 、P a+1 , kappa= |theta 2 −θ 1 |/deltat, wherein theta 1 、θ 2 is the azimuth angle of two adjacent displacement vectors, deltat is the time interval, T_gps=3 m when kappa >0.15rad/s, T_gps=8 m when kappa <0.03rad/s, and T_gps is linearly interpolated within the range of 3m to 8m in the middle curvature interval.

Description

Intelligent recording method for motion data Technical Field The invention relates to the technical field of motion data processing, in particular to a method and a system for diluting motion data intelligent recording and feature reservation, which are suitable for equipment such as mobile phones, servers and the like with strong computing power and abundant resources, and are also suitable for portable motion equipment with general computing power and limited resources. Background With the development of sports health industry, sports watches, smart bracelets, riding stopwatch and other devices can acquire multidimensional sports data such as GPS tracks, heart rate, speed, pedaling frequency, power and the like. However, high frequency acquisition results in a huge amount of data, not only occupying storage space, but also increasing transmission and analysis costs. The traditional data thinning method (such as equidistant sampling and fixed threshold filtering) does not consider the characteristic relevance of the motion data, and key information (such as a power peak value and curve track inflection point during sprint) is easy to lose, so that the motion analysis accuracy is affected. Most of the current motion data thinning technologies have the following defects: The correlation among data is ignored by single-dimensional thinning, and corresponding treading frequency/power data is not synchronously reserved when a speed mutation point is deleted; The fixed threshold cannot be adapted to different exercise intensities (such as jogging and sprint requiring different thinning strategies); The algorithm complexity is too high (such as high-order filtering and global optimization) to be operated on low-power equipment such as sports watches; feature extraction relies on global data (such as a complete track) and cannot be collected and processed at the same time; Device resource limitations (such as CPU, storage and power consumption) are not considered, and equipment is easy to be blocked; Aiming at the defects, the invention provides an intelligent thinning scheme which combines data compression and feature preservation and can smoothly run on high-calculation-force equipment such as a mobile phone/server and low-calculation-force equipment such as a sports bracelet/watch. Disclosure of Invention The problems of the prior art solved by the application are as follows: Aiming at the technical defects of the existing motion data thinning method in the aspects of insufficient feature reservation, low compression efficiency, high algorithm complexity, incapability of real-time processing, difficulty in adapting to resource-limited equipment and the like, the invention provides an intelligent recording method and system of motion data, according to the method, by introducing a multidimensional collaborative judgment mechanism, a dynamic threshold adjustment strategy, a sliding window buffer structure and a priority processing flow, the data storage and transmission overhead is obviously reduced, and meanwhile, key semantic feature information in a motion process is efficiently reserved, so that the method is applicable to various hardware platforms from low-power wearable equipment to a high-performance server. The technical scheme for solving the technical problems is as follows: An intelligent recording method for motion data is provided, which comprises the following steps, S1, carrying out time synchronization and denoising pretreatment on original motion data from a multi-source sensor; S2, dividing the preprocessed multidimensional motion data into a plurality of overlapped sliding windows according to time sequence, and caching the sliding windows in a memory; s3, executing multidimensional feature judgment on the data points except the head and the tail in each sliding window according to preset priorities in sequence, and marking the current point and terminating judgment of other dimensions once any dimension judges the current point as a feature point; s4, forcedly reserving a first group of data points and a last group of data points of each sliding window as characteristic points; And S5, only writing all the data marked as the feature points into a local storage or uploading the data to a remote server, and discarding the rest of original data. Preferably, in the step S1, the time synchronization process interpolates and aligns the heart rate, the speed, the pedal frequency, the power and the altitude data with reference to the time stamp output by the GPS module; The denoising treatment adopts first-order low-pass filtering, and the cut-off frequency is set to be 0.5Hz to 2Hz. Preferably, in the step S2, the sliding window is formed by n sets of multi-dimensional data aligned in time, and each set of data includes values of six dimensions of longitude and latitude, heart rate, speed, treading frequency, power and altitude of the GPS at the same time; Overlapping of one data point exists betwee