TinyML PIR Human Sensor for Low-False-Alarm Motion Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion detection systems, particularly passive infrared (PIR) sensors, suffer from high false alarm rates due to their inability to discriminate between animate and inanimate objects, leading to unnecessary resource allocation and costs.
Innovation Solution
A low-power passive infrared human sensor utilizing machine learning, specifically a TinyML model, is employed to classify movement sources as humans, adults, children, animals, or inanimate objects by analyzing infrared data with a convolutional neural network (CNN) on a microcontroller, reducing false alarms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional PIR sensors are used for motion detection, then the system is simple and low cost, but the false alarm rate is high due to inability to discriminate between animate and inanimate objects
Solution Approach 1:
The patent segments the motion detection task into multiple stages: initial motion detection by PIR sensor, followed by classification of motion source type (human, animal, inanimate) using machine learning models. This segmentation allows the simple PIR sensor to remain while adding intelligence only where needed to reduce false alarms.
Solution Approach 2:
The patent introduces machine learning models (TinyML and CNN) as intermediaries between the PIR sensor and the alarm trigger. These intermediaries analyze the motion patterns and classify the source, acting as a smart filter that prevents false alarms while maintaining system reliability.
2Reliability
If machine learning models are added to classify motion sources, then false alarm reduction is achieved, but power consumption increases
Solution Approach 1:
The system uses periodic action by implementing sleep-wake cycles for the microcontroller and using event-triggered inference. The machine learning models are only activated when motion is detected by the PIR sensor, rather than running continuously, significantly reducing power consumption while maintaining detection accuracy.
Solution Approach 2:
The patent employs parameter changes by using different machine learning model complexities (TinyML for low power, CNN for higher accuracy) and adjusting inference frequency based on motion patterns. This allows the system to optimize the balance between power consumption and detection accuracy dynamically.
3Productivity
If comprehensive motion classification is implemented, then resource optimization is improved, but processing time and computational load increase
Solution Approach 1:
The system performs preliminary action by using the PIR sensor to detect motion first, then only initiating machine learning classification when motion is present. This preliminary detection step filters out non-motion events before they reach the computationally intensive classification stage, reducing overall processing time.
Solution Approach 2:
The patent applies partial action by implementing tiered classification: basic motion detection for all events, then selective detailed classification only for detected motion events. This partial approach to classification reduces computational load while still achieving the resource optimization goal for relevant events.
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 accurately discriminates movement sources, minimizing false alarms and optimizing resource utilization by classifying humans and activities, thus enhancing energy efficiency and reducing operational costs.
Implementation Method 1
One or more passive infrared (IR) sensors can be mounted to receive infrared (IR) energy through a lens
Data Source
AI summary
Techniques for machine learning-based motion detection are disclosed. One or more passive infrared (IR) sensors are accessed and are coupled to a microcontroller for analysis. The one or more passive infrared sensors are mounted to receive light through a lens. IR data is collected from the one or more passive infrared sensors. The IR data is sampled at a sampling rate by the microcontroller. Movement is detected based on the collected IR data from the one or more passive infrared sensors. IR data is sent to a machine learning model. The machine learning model is based on a TinyML model. The TinyML model operates on the microcontroller. The TinyML model classifies the one or more animate and inanimate sources of the movement. Sources of movement include humans and human activity including adults and children.