Personalized Gesture Recognition via Sensor Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing gesture recognition systems in devices are not highly configurable, leading to inaccurate recognition due to differences in user performance and lack of adaptability to individual preferences, sensor changes, and non-portability of gesture settings across devices.
Innovation Solution
A method that allows users to configure and customize gesture recognition by monitoring sensor outputs during performance, identifying specific patterns, and associating actions with gestures, enabling personalized gesture definitions and adaptability across devices with varying sensor sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a device uses preprogrammed gesture recognition with model sensor outputs, then the device can recognize gestures performed in a typical manner, but the recognition accuracy decreases for gestures performed differently by individual users
Solution Approach 1:
The system performs preliminary training by monitoring sensor outputs while the user performs gestures before actual recognition. This preliminary action captures the user's specific gesture patterns and stores them as reference data, enabling accurate recognition of that user's gestures without requiring perfect model sensor outputs for all possible gesture variations.
Solution Approach 2:
The gesture recognition system serves itself by automatically adapting to each user through training. The system monitors its own sensor outputs during user gestures and uses this self-collected data to improve recognition accuracy, eliminating the need for external calibration or manual configuration.
2Adaptability or versatility
If the device uses a broad range of sensor outputs for gesture recognition, then more gesture variations are recognized, but false positives increase where non-gestural changes are incorrectly interpreted as gestures
Solution Approach 1:
The system applies local quality by creating user-specific gesture profiles that capture the precise characteristics of each user's gestures. Instead of using a single broad model for all users, the system tailors the recognition criteria to match each user's local gesture patterns, thereby recognizing genuine gesture variations while excluding non-gestural movements that don't match the user's specific pattern.
Solution Approach 2:
The training process incorporates feedback by monitoring sensor outputs during user-performed gestures and using this information to refine recognition criteria. The system learns from actual user behavior and adjusts its sensitivity thresholds based on observed patterns, reducing false positives while maintaining recognition of legitimate gesture variations.
3Device complexity
If gesture recognition settings are fixed in a device, then the configuration is simple, but the settings cannot be ported across multiple devices or adapted to sensor changes
Solution Approach 1:
The system creates portable copies of gesture profiles that can be transferred between devices. By storing user-specific sensor output patterns as copyable data structures, the system enables gesture recognition settings to be replicated across multiple devices, maintaining consistency of user experience without requiring complex configuration procedures on each device.
Solution Approach 2:
The gesture profile system achieves universality by designing a portable data structure that works across different devices and sensor configurations. The same gesture profile can be universally applied to multiple devices, and the system adapts to different sensor sets by leveraging the transferred profile data, eliminating the need for device-specific configuration.
4Device complexity
If the device uses static training to detect typical gesture examples, then the recognition process is simple, but it produces both false negatives and false positives
Solution Approach 1:
The system performs preliminary training by monitoring sensor outputs while the user performs gestures before actual recognition. This preliminary action captures the user's specific gesture patterns and stores them as reference data, enabling accurate recognition of that user's gestures without requiring perfect model sensor outputs for all possible gesture variations.
Solution Approach 2:
The system changes parameters by transitioning from static model sensor outputs to dynamic user-specific gesture profiles. By capturing actual sensor output data during user gestures and using this empirical data as the recognition basis, the system adapts parameters to match real user behavior, thereby eliminating both false negatives and false positives while maintaining process simplicity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In many computing scenarios, a device comprises at least one sensor, and is configured to recognize a gesture performed by a user according to the sensor output of the sensor, and to perform a particular action upon recognizing the gesture. However, many devices are preconfigured with such gestures, and the recognition is specific to the sensors of the device, and is not specific to the manner in which a particular user performs the gesture. Presented herein are techniques for enabling a device to recognize a new gesture by monitoring the sensor output of any sensors provided by the device while the user performs the gesture, optionally requesting repeated gesture performances until reaching a recognition confidence. Once trained to recognize the gesture according to the sensor outputs of the particular sensors of the device, the device may subsequently recognize the gesture performed by the user and execute an associated action.