Gesture Recognition Using HMMs and Lookup Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing gesture recognition systems in three-dimensional spaces are prone to errors due to uncertainties introduced in processing stages, particularly when using x/y/z estimates, and are not invariant to translation and scaling, affecting reliability.
Innovation Solution
A method and system that uses Hidden Markov Models (HMMs) for gesture recognition, employing features like non-quantized or quantized sensor signal levels, position, distances, direction, and derivatives, with start and stop criteria based on distance variance and signal changes, to detect gestures invariant to translation and scaling, and applicable to capacitive and other sensor technologies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If x/y/z position estimates are used as input to the gesture recognition system, then the gesture detection can be performed using straightforward high-level approach, but the recognition reliability deteriorates due to uncertainties introduced in processing stages
Solution Approach 1:
The patent applies preliminary action by performing calibration and creating lookup tables during an initialization phase before actual gesture recognition. The system pre-computes the relationship between sensor signals and spatial positions, storing this mapping in lookup tables. During runtime, gesture recognition directly queries these pre-computed tables rather than performing complex real-time calculations, thereby eliminating processing uncertainties while maintaining ease of operation.
2Measurement precision
If calibration and non-linear relation processing stages are introduced to improve measurement accuracy, then the position estimation improves, but the system complexity and error propagation increase
Solution Approach 1:
The patent creates simplified copies of the complex non-linear relationships by pre-computing lookup tables that represent the calibrated sensor signal to position mapping. Instead of implementing complex non-linear processing algorithms during runtime, the system copies the essential relationship information into tabular form that can be queried directly. This reduces computational complexity while preserving measurement precision.
3Reliability
If the gesture recognition system is made invariant to translation and scaling, then the system becomes more robust to variations in gesture performance, but the processing complexity increases
Solution Approach 1:
The patent achieves translation and scaling invariance by changing the reference frame parameters. Instead of processing absolute position coordinates that vary with translation and scaling, the system transforms sensor signals into relative position relationships and uses calibrated lookup tables that are invariant to these transformations. This allows the system to recognize gestures based on their intrinsic geometric relationships rather than absolute positions, providing robustness without excessive complexity.
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
The system achieves reliable and robust gesture recognition with high accuracy, demonstrated by recognition rates of at least 95% across different users, effectively addressing the issues of error-proneness and variability in gesture performance.
Implementation Method 1
Capacitive sensor systems, for example, can be realized by generating an alternating electrical field and measuring the potential difference (i.e., the voltage) obtained at a sensor electrode within the field. The voltage at the sensor electrode(s) is a measure for the capacitance between the sensor electrode and its electrical environment.
Implementation Method 2
Capacitive sensor systems, for example, can be realized by generating an alternating electrical field and measuring the potential difference (i.e., the voltage) obtained at a sensor electrode within the field.
Data Source
Figure 1
Figure 2~3
Figure 4~5B
AI summary
A method for gesture recognition including detecting one or more gesture-related signals using the associated plurality of detection sensors; and evaluating a gesture detected from the one or more gesture-related signals using an automatic recognition technique to determine if the gesture corresponds to one of a predetermined set of gestures.