Tag Device Motion Recognition via Sensor Fusion and Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion recognition technologies face challenges in achieving high accuracy while minimizing computational requirements, particularly due to environmental factors and low accuracy in non-optical methods.
Innovation Solution
A tag device that communicates with anchor devices, utilizing a first sensor for position data and a second sensor for speed data, generates final position data to recognize motion through a neural network model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical method (camera and image sensor) is used for motion recognition, then motion detection capability is improved, but calculation amount increases and accuracy is affected by environmental factors
Solution Approach 1:
The patent segments the motion recognition process into two distinct parts: (1) position data acquisition using sensors with low computational requirements, and (2) motion pattern recognition using neural networks. This segmentation allows each component to be optimized independently, reducing overall system complexity while maintaining accuracy.
Solution Approach 2:
The patent introduces an intermediary representation - converting sensor data into position information and then into motion patterns before neural network processing. This intermediary step simplifies the data structure and reduces the computational burden on the neural network while preserving essential motion characteristics.
2Device complexity
If non-optical method (IMU sensor) is used for motion recognition, then calculation amount is reduced, but motion recognition accuracy becomes low
Solution Approach 1:
The patent merges multiple sensor types (first sensor for position data, second sensor for speed data, third sensor for additional position data) to compensate for the limitations of individual sensors. By combining data from multiple low-complexity sensors, the system achieves high motion recognition accuracy without requiring complex optical methods.
Solution Approach 2:
The patent transforms raw sensor parameters (acceleration, angular velocity) into derived parameters (position, speed, motion patterns) through integration and neural network processing. This parameter transformation enables accurate motion recognition using simple sensors by changing the data representation rather than the sensing method.
3Measurement precision
If multiple sensors are used to improve motion recognition accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent designs a unified processing architecture that handles data from multiple sensor types through a common pipeline: data acquisition → position calculation → motion pattern extraction → neural network recognition. This universal processing approach manages multi-sensor complexity by treating different sensors through a single integrated framework.
Data Source
AI summary
There is provided a tag device including a first sensor, a second sensor, a pre-processor and a neural network processor. The pre-processor generates first position data of the tag device based on time information sensed by at least one of a third sensor included in each of the one or more anchor devices and the first sensor of the tag device, generates second position data based on first speed data of the tag device sensed by the second sensor and the first position data, and generates an image based on a path of movement of the tag device based on the second position data in an operation period, and the neural network processor classifies the image into one of a plurality of movements by using a trained neural network model.