Dynamic Threshold Motion Detection Using Curve Fit Error Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Detecting human motion using ambient motion sensors is challenging due to noise sources from external and internal environments, such as airflow, electromagnetic interference, electrostatic discharge, and radio frequency currents, which complicate the differentiation between human movement and noise signals.
Innovation Solution
A method and system that generate a motion event signal, fit a parameterized curve to sensor data to filter noise, estimate noise source signal magnitudes, and dynamically adjust detection thresholds based on curve fit errors and noise estimates to accurately detect human motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If ambient motion sensors are used for human motion detection, then device cost and size are reduced, but detection accuracy deteriorates due to noise from external and internal sources
Solution Approach 1:
The system performs preliminary characterization of noise sources before actual motion detection. Noise profiles from various sources (HVAC, EMI, ESD, RF) are captured and stored in advance, allowing the filter to recognize and compensate for these known interference patterns during operation, thereby improving detection accuracy with low-cost sensors
Solution Approach 2:
A digital signal processing filter acts as an intermediary between the noisy sensor output and the motion detection decision. The filter uses parameterized curves to model and subtract noise components, effectively mediating the interference between noise sources and detection accuracy while allowing the use of inexpensive sensors
2Measurement precision
If motion detection threshold is set low to detect subtle human motion, then detection sensitivity is improved, but false positives increase due to noise signals
Solution Approach 1:
The system uses feedback from the parameterized curve fitting process to dynamically adjust detection thresholds. The curve fit error provides continuous feedback about the presence of noise, allowing the system to raise thresholds when noise is detected and lower them when the signal is clean, thereby maintaining high sensitivity while reducing false positives
Solution Approach 2:
The detection threshold is not fixed but dynamically changed based on the curve fit error and noise estimates. When the parameterized curve poorly fits the sensor data (indicating noise), the threshold is adjusted upward; when the fit is good, the threshold can be lower, enabling flexible adaptation to varying noise conditions
3Reliability
If motion detection threshold is set high to reduce false positives, then reliability is improved, but detection sensitivity deteriorates and false negatives increase
Solution Approach 1:
The curve fit error provides feedback that prevents excessive threshold elevation. Even when noise is present, the system can determine how much of the signal is attributable to noise versus actual motion, allowing it to maintain lower thresholds with confidence by compensating for the known noise components through the parameterized modeling
4Device complexity
If fixed detection threshold is used for motion detection, then device complexity is reduced, but adaptability to varying noise conditions deteriorates
Solution Approach 1:
The detection threshold transitions from a fixed value to a dynamic parameter that adapts to changing noise conditions. The system continuously updates the threshold based on real-time curve fit errors and noise estimates, allowing the threshold to move appropriately as noise levels change, thereby achieving adaptability with moderate complexity
Data Source
AI summary
Systems and methods of detecting human movement with a sensor are provided, including generating a motion event signal in response to movement detected by the sensor, and generating a parameterized curve to represent the detected motion. The parameterized curve is fit to a predetermined window of sensor data captured by the sensor to filter the motion event signal. A noise magnitude estimate and a curve fit error is determined based on the fitted parameterized curve to the predetermined window. A detection threshold value is determined based on the curve fit error, a noise source signal estimate of known noise, and zero or more noise magnitudes from other sources. Human motion is determined by correlating a true motion event signal with human motion based on a comparison between a value of a point on the parameterized curve and the detection threshold value.