Gesture Inference ML Model Selection for Session Variability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biopotential gesture machine interfaces face challenges due to inter/intra-session variability, making gesture inference difficult and inaccurate, and integrating data from disparate sources such as cameras and biopotential sensors complicates the process.
Innovation Solution
A system and method that combines camera-based computer vision with biopotential sensing wearable devices, using multiple machine learning models to process and synchronize data, applying transformations to address variability and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If biopotential gesture machine interfaces are used to classify user input, then gesture recognition capability is provided, but inter/intra-session variability makes inference difficult and reduces accuracy
Solution Approach 1:
The patent combines multiple data sources including biopotential sensor data, motion sensor data, and camera-based computer vision data into a unified gesture recognition system. This multi-modal fusion approach compensates for the limitations of individual sensors by leveraging complementary information from different sources, thereby improving inference reliability and accuracy despite inter-session variability.
Solution Approach 2:
The system dynamically adjusts ML model parameters and selection based on session characteristics and data quality. By changing parameters such as model architecture, hyperparameters, and data weighting schemes according to session conditions, the system adapts to variability while maintaining high classification accuracy across different usage scenarios.
2Adaptability or versatility
If multiple data sources (cameras and biopotential sensors) are integrated, then comprehensive gesture detection is achieved, but data handling complexity increases
Solution Approach 1:
The patent divides the complex data processing task into separate specialized modules: one module handles biopotential sensor data processing, another handles motion sensor data, and a third handles computer vision data. Each module independently processes its specific data type using appropriate algorithms, then the results are fused at a higher level. This segmentation reduces overall system complexity by making each component more manageable and easier to optimize independently.
Solution Approach 2:
The system introduces an intermediary layer that standardizes and synchronizes data from different sources before processing. This intermediary module handles timestamp alignment, data format normalization, and coordination between disparate data streams, thereby simplifying the integration process and reducing the complexity of handling multiple data sources with different characteristics.
3Reliability
If ML model transformations are applied to address session variability, then gesture inference robustness is improved, but processing time and computational overhead increase
Solution Approach 1:
The system performs preliminary actions by pre-processing and transforming ML model inputs before the main inference process. Data normalization, feature extraction, and initial filtering are conducted in advance to reduce the computational burden during actual gesture recognition. This preliminary preparation reduces processing time during critical inference moments while maintaining robustness through proper data conditioning.
Solution Approach 2:
The patent implements dynamic model selection and transformation strategies that adapt processing intensity based on session conditions. When session variability is low, simpler models with faster processing are used. When variability increases, more robust transformations are applied. This dynamic approach balances robustness and processing time by adjusting the level of transformation based on real-time conditions rather than always applying maximum processing.
Data Source
AI summary
Disclosed are methods, systems and non-transitory computer readable memory for gesture inference. For instance, a first method may include computer vision to train and/or infer gesture inferences. For instance, a second method may include using transformations to data and/or ML models to address inter/intra-session variability of sensor data. For instance, a third method may include using ML model selection to select a ML model to address inter/intra-session variability of sensor data.