Wearable Device Smoking Detection via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current smoking cessation methods rely on self-reporting, which is unreliable, and lack mobile device-based solutions for the increasingly mobile population, making it difficult for smokers to track and manage their smoking habits effectively.
Innovation Solution
A wearable computing device equipped with sensors and a machine-learned smoking gesture detection model, trained to detect smoking gestures through sensor data, providing accurate tracking and alerts to users, regardless of device orientation, and integrating with a smartphone for summary statistics and encouragement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If self-reporting methods are used for smoking tracking, then users can report their smoking habits, but the data reliability is poor due to user trust issues and inconsistency
Solution Approach 1:
The system enables automatic detection and tracking of smoking behaviors through machine learning models that analyze sensor data from the wearable device, eliminating the need for users to manually self-report their smoking habits while ensuring consistent and reliable data collection
Solution Approach 2:
The patent replaces the mechanical/manual self-reporting system with an automated electronic detection system using machine learning algorithms that process accelerometer and other sensor data to automatically identify and record smoking events
2Adaptability or versatility
If traditional smoking cessation websites are used, then resources and support are available, but they lack mobile device integration for the mobile population
Solution Approach 1:
The wearable device integrates multiple functions including smoking detection, health monitoring, and mobile device connectivity into a single platform, making the smoking cessation solution accessible and adaptable to the mobile population through smartphone integration and wireless communication
3Measurement precision
If simple smoking tracking apps are used, then users can input smoking data, but the system complexity is low and cannot provide accurate real-time detection
Solution Approach 1:
The patent introduces machine learning models as intermediary components that process raw sensor data from the wearable device, transforming complex accelerometer signals into accurate smoking event detections through trained algorithms that recognize smoking-specific motion patterns
Data Source
AI summary
The present disclosure provides a wearable computing device. The wearable computing device includes one or more sensors that output sensor data. The wearable computing device includes a machine-learned smoking gesture detection model trained to detect a smoking gesture based on the sensor data. The wearable computing device includes one or more processors and one or more non-transitory computer-readable media that store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations. The operations include inputting the sensor data into the machine-learned smoking gesture detection model. The operations include receiving, as an output of the machine-learned smoking gesture detection model, a gesture classification that indicates whether the sensor data is indicative of the smoking gesture. The operations include determining whether a user of the wearable computing device is engaging in a smoking session based at least in part on the gesture classification.