Gait Motion Recognition Using Hip Angle and Acceleration Data

The patent segments the gait motion recognition task into multiple classification stages using a hierarchical approach. First, it distinguishes between downward gait motions (stairs/slope) and other motions using vertical acceleration thresholds. Then, it uses neural networks to differentiate between level walking, upward walking, and downward slope walking. This segmentation allows the system to handle complex gait variations through a structured, multi-level decision process rather than a single complex model.

The patent implements dynamic adaptation by adjusting the recognition process based on real-time sensor data. The system dynamically selects which classification method to apply (threshold-based vs. neural network-based) depending on the detected vertical acceleration characteristics. This dynamic approach enables the system to optimize its behavior for different gait conditions without requiring a completely different system for each scenario.

The patent divides the gait recognition task into two segments: a simple threshold-based initial classification for downward gait motions (using vertical acceleration), and a more complex neural network-based classification for other motions. This segmentation allows the system to use simple, easy-to-operate threshold checks for clear-cut cases while reserving complex neural network analysis for situations requiring finer differentiation, thus balancing ease of operation with measurement precision.

The patent changes the recognition parameters dynamically based on the detected motion characteristics. For downward gait motions, it uses vertical acceleration thresholds as the primary parameter. For other motions, it switches to using neural network processing of hip joint angle and acceleration data. This parameter adaptation allows the system to optimize between simplicity and precision based on the specific gait condition being detected.

The patent applies segmentation by using neural networks only for specific classification tasks where they are most needed (differentiating level walking, upward walking, and downward slope walking), while using simpler threshold-based methods for initial downward gait detection. This selective application of neural networks reduces the overall system complexity compared to using neural networks for all gait motion recognition, while still maintaining high measurement precision for the critical differentiation tasks.

The patent applies partial action by using neural networks only for the portion of gait recognition that requires sophisticated pattern analysis (differentiating similar gait types), rather than applying them universally to all recognition tasks. This partial application of complex processing where needed most optimizes the balance between measurement precision and device complexity.

The patent segments the data processing workflow into stages: first collecting comprehensive sensor data (hip joint angles, vertical acceleration, rotational speed), then applying simple threshold checks to filter and pre-classify data, and finally using neural networks only for the remaining classification tasks. This segmentation reduces the energy consumption compared to continuously processing all data through complex neural networks, while still achieving high measurement precision through the multi-stage approach.

The patent applies partial processing by using intensive neural network analysis only for the portion of data that requires sophisticated interpretation (gait type differentiation), while using lighter threshold-based processing for initial classification. This partial application of computationally intensive processing reduces overall energy consumption while maintaining the measurement precision needed for accurate gait recognition.