Targeted Cough Detection With Multi-Stage User Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cough detection technologies face limitations due to reliance on speech datasets, model dependency issues, and challenges in converging during training, particularly with triplet loss functions, leading to inefficiencies and inflexibility in feature engineering and computational complexity.
Innovation Solution
A multi-stage cough detection system utilizing edge and cloud detectors, employing convolutional neural networks and Transformer models, with environment-specific training datasets to isolate and filter cough sounds from non-cough sounds, ensuring privacy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If unobtrusive remote sensing is used to monitor coughs, then patient monitoring convenience is improved, but the system cannot distinguish between the targeted user's coughs and other people's coughs
Solution Approach 1:
The system segments the audio monitoring task into two distinct stages: first, a general cough detector identifies potential cough sounds in the environment; second, a targeted cough detector specifically verifies whether the detected cough belongs to the targeted user by comparing acoustic features against stored reference profiles. This segmentation allows the system to maintain unobtrusive monitoring while achieving accurate user-specific identification.
Solution Approach 2:
The patent introduces an intermediary verification mechanism between general cough detection and final diagnosis. The targeted cough detector acts as an intermediary that filters and validates cough detections by comparing them against user-specific reference profiles, thereby ensuring that only authenticated user coughs are used for health monitoring without requiring direct user interaction.
2Ease of manufacture
If speech datasets are used for training cough detection models, then model training is simplified, but the feature engineering is not optimal for capturing cough characteristics
Solution Approach 1:
The system changes the parameter set used for feature extraction by implementing cough-specific acoustic feature engineering that goes beyond standard speech features. This includes extracting features specifically relevant to cough acoustics such as abrupt onset detection, specific frequency range analysis, and temporal pattern recognition, thereby optimizing the model for cough detection while maintaining practical trainability.
3Measurement precision
If triplet loss function is used for training embedding models, then user-specific cough identification capability is improved, but training convergence becomes difficult and computational complexity increases
Solution Approach 1:
The system performs preliminary action by pre-computing and storing reference cough profiles for each user before actual monitoring begins. During operation, the targeted cough detector compares incoming cough detections against these pre-established references using efficient similarity metrics, avoiding the need for complex real-time triplet loss computation and enabling fast, accurate user identification without ongoing training complexity.
Data Source
AI summary
A method to detect cough of a targeted user includes processing a cough audio clip with a targeted cough detector and generating based thereon a result indicative of whether the cough audio clip contains a cough sound of the targeted user.