Radar Gesture Recognition via Shapelet Decomposition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional radar-based gesture recognition systems face challenges in accurately distinguishing different gestures due to their reliance on complex radar systems and machine learning techniques, which result in low classification accuracy and are highly dependent on collected data, especially for non-rigid subjects like humans.

Innovation Solution

The implementation of a shapelet decomposition method using a radar system with three vertically arranged radars, where time domain signals are processed to extract and decompose shapelets into superimposed components, allowing for accurate gesture recognition by segregating limb movements and applying signal processing techniques like spectrogram analysis and time delay adjustments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional radar systems (UWB, FMCW) and machine learning techniques are used for gesture recognition, then the system can detect gestures, but the classification accuracy is low and the system becomes highly dependent on collected data

Engineering Contradiction:
Improvegesture classification accuracyVSAvoidradar system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the gesture recognition problem into distinct components: (1) dividing the radar detection space into multiple regions using multiple radars, (2) separating different gestures into distinct clusters using clustering algorithms, and (3) decomposing complex gestures into basic gesture components. This segmentation approach improves classification accuracy by making each recognition task more manageable and less dependent on large datasets.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes key parameters of the radar system including operating frequency, pulse width, and power levels to optimize gesture detection. By adjusting these parameters, the system achieves better signal-to-noise ratio and gesture discrimination capability without requiring overly complex hardware configurations.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If machine learning techniques are used for feature extraction from time domain signals, then the system can process gestures, but it fails to effectively distinguish different gestures

Engineering Contradiction:
Improvegesture distinction accuracyVSAvoiddependency on collected data
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent extracts essential gesture features by applying clustering algorithms to identify and separate distinct gesture patterns from radar signals. This extraction process isolates the most discriminative characteristics of each gesture, enabling effective distinction without requiring extensive training data for machine learning models.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent performs preliminary clustering and classification of gestures before final recognition. By pre-processing the signals to group similar gestures together and identify basic gesture components beforehand, the system reduces the complexity of the main recognition task and minimizes dependency on large collected datasets for training.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If a single radar is used for gesture detection, then the system is simple, but it cannot accurately track movement and resolve orientational ambiguity

Engineering Contradiction:
Improvemovement tracking accuracyVSAvoidnumber of radars
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines data from multiple radars to achieve accurate movement tracking and gesture recognition. By merging the detection results from multiple radar units positioned at different locations, the system resolves orientational ambiguity and achieves comprehensive three-dimensional gesture monitoring without requiring an overly complex individual radar design.

Inventive Principle:
Principle #5Merging (Combining)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach enhances gesture recognition accuracy by addressing variations in displacement and speed across different individuals, providing distinct shapelets for each gesture, thus improving classification precision and reducing dependency on collected data.

Implementation Method 1

obtain, using a radar system, a plurality of time domain signals reflected by a subject performing one or more gestures

Methodology Applied
Scientific EffectRadar: Radar

Implementation Method 2

obtain, using a processor, a plurality of spectrograms corresponding to the plurality of time domain signals

Methodology Applied
Scientific EffectSpectrogram analysis:

Data Source

PatentEP3839811B1Systems and methods for shapelet decomposition based gesture recognition using radar
Publication Date: 2024.03.13 TATA CONSULTANCY SERVICES LTD
  • EP3839811B1 patent drawingFigure 1
  • EP3839811B1 patent drawingFigure 2
  • EP3839811B1 patent drawingFigure 3

AI summary

This disclosure relates to systems and methods for shapelet decomposition based recognition using radar. State-of-the-art solutions involve use of standard machine learning classification techniques for gesture recognition which suffer with problem of dependency on collected data. The present disclosure overcome the limitations faced by the state-of-the-art solutions by obtaining a plurality of time domain signal using a radar system comprising three vertically arranged radars and one or more sensors, identifying one or more gesture windows to obtain one or more shapelets corresponding to one or gestures which are further decomposed into a plurality of sub shapelets. Further, at least one of (i) a positive or (i) a negative time delay is applied to each of the plurality of sub shapelets to obtain a plurality of composite shapelets which are further mapped with a plurality of trained shapelets to recognize gestures comprised in one or more activities performed by a subject.