Wrist-Tilt Gesture Detection Using Adaptive Accelerometer Streams
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wrist-tilt gesture detection methods in wearable electronic devices suffer from limited performance in terms of latency and accuracy, especially when the device is subjected to high accelerations, such as during running.
Innovation Solution
The method involves receiving an input stream of samples from an accelerometer constrained to the user's wrist, executing an automatic-learning algorithm to identify conditions of movement, generating multiple streams of samples, and selecting the appropriate stream for wrist-tilt gesture detection based on the identified condition of movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If accelerometer-based wrist-tilt detection is used, then power consumption is reduced, but detection accuracy and performance deteriorate under high accelerations
Solution Approach 1:
The system dynamically adapts the detection algorithm based on detected movement conditions. When high acceleration is detected (e.g., during running), the system switches to a different detection mode that accounts for the dynamic conditions, thereby maintaining accuracy without continuously using high-power sensors like gyroscopes.
Solution Approach 2:
The system changes detection parameters (thresholds, filtering levels, algorithm selection) based on the detected movement state. By monitoring acceleration magnitudes and adjusting detection parameters accordingly, the system optimizes accuracy for different activity levels while maintaining low power consumption during low-activity states.
2Device complexity
If accelerometer-based wrist-tilt detection is used, then device complexity is reduced, but detection latency increases
Solution Approach 1:
The system performs preliminary analysis of acceleration patterns to predict upcoming gestures. By detecting preparatory movements and pre-processing the data stream, the system can reduce latency when the actual wrist-tilt gesture occurs, without adding complex hardware.
Solution Approach 2:
During periods of detected high activity or when gesture patterns are clearly emerging, the system skips unnecessary processing steps or reduces filtering intensity to accelerate detection, thereby reducing latency while maintaining accuracy through subsequent verification.
3Device complexity
If standard threshold-based detection is used, then device complexity is minimized, but reliability deteriorates under varying movement conditions
Solution Approach 1:
The system dynamically adjusts detection thresholds and parameters based on the current movement context. By continuously monitoring acceleration patterns and adapting the detection criteria, the system maintains high reliability across varying activities (walking, running, stationary) without requiring complex machine learning models.
Solution Approach 2:
The system uses feedback from continuous acceleration monitoring to adjust detection parameters in real-time. By analyzing the reliability of current detection conditions and adapting thresholds accordingly, the system maintains consistent performance across different movement scenarios while keeping the algorithm relatively simple.
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 enables low latency and high accuracy in wrist-tilt gesture detection across various user movements, including running, without the need for high-power consumption components like gyroscopes.
Implementation Method 1
receiving an input stream of samples from an accelerometer constrained to an user's wrist, the samples of said input stream being indicative of acceleration values along at least one axis
Implementation Method 2
executing an automatic-learning algorithm on blocks of samples of the input stream of samples to identify, for each block, a corresponding condition of movement of the user
Implementation Method 3
generating a filtered stream of samples by digital low-pass filtering of samples of at least part of the input stream of samples
Data Source
AI summary
An electronic device has an input which, in operation, receives an input stream of accelerometer data samples indicative of acceleration values along at least one axis. The devices includes circuitry, coupled to the input. The circuitry, in operation, executes an automatic-learning algorithm on blocks of samples of the input stream of accelerometer data samples to identify, for each block, a corresponding condition-of-user-movement from among a plurality of determined conditions-of-user-movement. The circuitry generates a plurality of streams of samples based on the input stream of accelerometer data samples, and for each condition of movement identified, selects a corresponding stream of samples of the plurality of streams of samples. The circuitry executes a wrist-tilt gesture detection algorithm using samples of the selected stream of the plurality of streams of samples.