Sensor Data Resampling for Storage-Limited Time Series Classification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The challenge of storing high-resolution time series data in electronic devices is limited by hardware storage capacity, leading to difficulties in accurate time series classification, particularly in sectors like healthcare where sensor storage cycles are affected.

Innovation Solution

An electronic device that generates and stores data using a combination of high and low sampling periods, interpolating and applying weights to reconstruct missing data points, enabling improved time series classification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If high-resolution time series data is stored continuously in the electronic device, then the accuracy of time series classification is improved, but the storage capacity of the device is exceeded

Engineering Contradiction:
Improveclassification accuracyVSAvoidstorage capacity
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent segments time series data into different sampling periods (first sampling period and second sampling period). High-resolution data is stored at a first sampling period while lower-resolution data is stored at a second sampling period, dividing the storage burden and enabling classification accuracy to be maintained for critical periods while reducing overall storage requirements.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the sampling period parameter of the data based on priority and time characteristics. By varying the sampling period (from first sampling period to second sampling period), the system optimizes storage usage while maintaining classification accuracy where needed, directly addressing the contradiction between storage capacity and classification performance.

Inventive Principle:
Principle #35Parameter changes

2Quantity of substance

If data is stored at a low sampling period to save storage space, then storage capacity is preserved, but the accuracy of time series classification deteriorates

Engineering Contradiction:
Improvestorage capacityVSAvoidclassification accuracy
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent applies local quality by assigning different data qualities (different sampling periods) to different time segments. Critical time periods are stored at the first sampling period with high resolution, while non-critical periods use the second sampling period with lower resolution. This ensures classification accuracy is maintained locally where needed while preserving overall storage capacity.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system performs preliminary classification to identify which time periods require high-resolution data before storage. By pre-determining which segments need the first sampling period versus the second sampling period, the system optimizes storage allocation in advance, ensuring classification accuracy is preserved for important segments without wasting storage capacity on less critical data.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4730165A1Electronic device for obtaining data having changed sampling period, and control method therefor
Publication Date: 2026.04.22 SAMSUNG ELECTRONICS CO LTD
  • EP4730165A1 patent drawingFigure 1
  • EP4730165A1 patent drawingFigure 2
  • EP4730165A1 patent drawingFigure 3A~3B

AI summary

An electronic device for obtaining data having a changed sampling period, and a control method therefor are disclosed. The electronic device according to one embodiment of the present document comprises at least one sensor, a memory and at least one processor, wherein the memory can include instructions that, when executed, cause the at least one processor to: generate second data including information about an attribute of data corresponding to a first time point in first data acquired during a first time interval through the at least one sensor in a first sampling period; interpolate the generated second data so as to generate third data on the basis of a second sampling period of which the sampling period differs from that of the first sampling period; and apply a weight to data corresponding to at least one time point to be reconstructed of the generated third data so as to generate fourth data on the basis of the second sampling period.