Cough Detection System Using ML for Asymptomatic Infection Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital questionnaires for monitoring health and social behaviors are unreliable in detecting asymptomatic individuals who can spread illnesses, as they rely solely on self-reported symptoms, missing those who do not exhibit symptoms.
Innovation Solution
A cough detection system that uses a trained machine learning model to analyze cough utterances from users, processing environmental and health-related factors to predict potential infections and notify individuals and authorities, enhancing the accuracy and reliability of infection control measures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If digital questionnaires rely solely on self-reported symptoms, then the system is simple and easy to operate, but the reliability of infection detection deteriorates because asymptomatic individuals cannot be identified
Solution Approach 1:
The patent replaces the mechanical system of self-reported questionnaires with an acoustic detection system using machine learning models. The system captures cough sounds through microphones, processes them through trained ML models to detect asymptomatic infections, and generates alerts. This substitution transforms the detection mechanism from subjective self-reporting to objective acoustic analysis, significantly improving reliability without requiring complex medical equipment.
Solution Approach 2:
The patent introduces cough sound analysis as an intermediary between symptom reporting and infection detection. Instead of directly relying on self-reported symptoms, the system uses acoustic signals of coughs as an intermediate indicator that can reveal asymptomatic infections. This intermediary layer enables indirect detection of infections that would otherwise go undetected by direct questioning alone.
2Measurement precision
If the system expands to detect asymptomatic carriers through cough analysis, then the measurement precision of infection detection improves, but the difficulty of detecting and measuring increases due to the need for machine learning models and environmental factor analysis
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models with extensive cough datasets before deployment. The models are trained in advance to recognize various cough patterns associated with different infections, including asymptomatic cases. This preliminary training phase stores learned patterns that enable rapid, accurate detection during actual use without requiring complex real-time analysis, thus reducing the operational difficulty while maintaining high precision.
Solution Approach 2:
The patent changes parameters by analyzing multiple environmental and contextual factors alongside cough sounds, such as location data, time of day, and user health information. By adjusting and weighting these parameters dynamically, the system improves detection precision while managing complexity through parameter optimization rather than requiring equally complex analytical frameworks for each factor.
3Productivity
If the system implements real-time cough monitoring and alerting, then the productivity of infection control improves through timely notifications, but the loss of time increases due to data processing and model inference requirements
Solution Approach 1:
The patent implements periodic action by analyzing cough sounds at regular intervals rather than requiring continuous monitoring. The system captures audio samples periodically, processes them through the ML model at scheduled times, and generates alerts based on accumulated data. This periodic approach maintains effective infection control while reducing computational burden and processing time compared to truly continuous real-time analysis, thus balancing productivity with time efficiency.
Data Source
AI summary
Methods and systems for a cough detection system are provided. The methods and systems include operations comprising: receiving an utterance comprising a cough from a user; determining one or more environmental factors, one or more health related factors, or both associated with the user; processing the utterance with a machine learning technique to generate an illness prediction, the machine learning technique being trained to establish a relationship between a plurality of training cough utterances and a plurality of illnesses corresponding to the plurality of training cough utterances; applying a weight to the illness prediction based on the one or more environmental factors, health related factors, or both associated with the user; and triggering an alert representing the illness prediction based on the weighted illness prediction.