Gesture Recognition via Dynamic Coordinate Alignment and Pre-Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current gesture recognition technologies face challenges such as dependence on device orientation, high false alarm rates, and computational complexity, limiting the usability and efficiency of motion gestures on mobile devices.
Innovation Solution
The method involves determining a rotation matrix between an input gesture and a template gesture to align them in the same coordinate system, applying filters to eliminate non-matching template gestures, and using a gesture recognition classifier like Dynamic Time Warping to improve recognition accuracy and reduce computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gesture recognition is performed using traditional methods comparing input gestures to template gestures in fixed coordinate systems, then gesture recognition can be implemented, but the recognition accuracy decreases when device orientation changes
Solution Approach 1:
The patent applies dynamics by making the coordinate system dynamic rather than fixed. The system determines device orientation using sensors (accelerometer, gyroscope, magnetometer) and dynamically transforms template gesture coordinate systems to match the current device orientation. This allows gesture templates to adapt to different device orientations in real-time, enabling accurate recognition regardless of how the device is held or oriented in space.
Solution Approach 2:
The patent changes the parameter of coordinate system orientation by using rotation matrices to transform template gestures from their original coordinate systems to the current device coordinate system. By adjusting the orientation parameters based on sensor data, the system maintains recognition accuracy across different device orientations without requiring the user to maintain a specific device posture.
2Measurement precision
If all template gestures are processed through gesture recognition classifiers to ensure accurate recognition, then recognition accuracy is maintained, but computational complexity and power consumption increase
Solution Approach 1:
The patent applies preliminary action by performing filtering operations before the full gesture recognition classification process. The system first compares basic gesture characteristics (such as direction, magnitude, or simplified features) against thresholds or reference values to eliminate obviously non-matching templates. Only templates that pass this preliminary filter undergo the computationally intensive classification process, significantly reducing overall power consumption while maintaining recognition accuracy for valid gestures.
Solution Approach 2:
The patent segments the gesture recognition process into multiple stages: a first filtering stage that quickly eliminates non-matching templates using simplified comparisons, and a second classification stage that applies full recognition algorithms only to promising candidates. This segmentation reduces the number of expensive computations performed on every template, lowering power consumption while preserving recognition accuracy through the multi-stage approach.
3Reliability
If computational resources are allocated to process all template gestures through classification, then gesture recognition completeness is maintained, but processing time and computational complexity increase
Solution Approach 1:
The patent performs preliminary filtering of template gestures before applying full classification algorithms. By first evaluating basic gesture properties against acceptance criteria, the system identifies and eliminates templates that cannot possibly match the input gesture. This preliminary action reduces the set of templates requiring complex computational processing, thereby reducing overall computational complexity while maintaining recognition completeness for valid gestures.
Solution Approach 2:
The patent divides the gesture recognition workflow into segmented stages: an initial filtering phase using simple comparisons to eliminate non-matching templates, followed by a secondary classification phase for remaining candidates. This segmentation reduces computational complexity by avoiding unnecessary processing of clearly non-matching templates, while the multi-stage approach ensures that potentially valid gestures receive thorough analysis for complete and reliable recognition.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and apparatuses are provided for facilitating gesture recognition. A method may include constructing a matrix based at least in part on an input gesture and a template gesture. The method may further include determining whether a relationship determined based at least in part on the constructed matrix satisfies a predefined threshold. In an instance in which the relationship does not satisfy the predefined threshold, the method may also include eliminating the template gesture from further consideration for recognition of the input gesture. In an instance in which the relationship satisfies the predefined threshold, the method may further include determining a rotation matrix based at least in part on the constructed matrix. Corresponding apparatuses are also provided.