Cough Detection via Acoustic and Visual Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cough detection devices inaccurately identify noise as cough sounds due to the presence of voiceless sound components, leading to false detections.
Innovation Solution
A cough detection system that combines acoustic feature extraction with image analysis, using a microphone array and camera to determine the arrival direction of sounds and identify coughing actions, thereby reducing false positives by correlating sound and image data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If acoustic data alone is used for cough detection, then the detection process is simple, but false detections increase due to noise misidentification
Solution Approach 1:
The patent combines acoustic data from a microphone array with image data from a camera to perform cough detection. The acoustic feature extractor processes sound data while the image processor analyzes visual data, and the determination unit integrates both results to identify coughs, thereby reducing false detections caused by noise alone
Solution Approach 2:
The determination unit acts as an intermediary that receives both acoustic determination results and image determination results, then integrates them to make the final cough detection decision. This mediator component reconciles the two different data sources and produces a more reliable outcome
2Productivity
If only acoustic features are analyzed, then processing is faster, but measurement precision decreases due to voiceless sound components
Solution Approach 1:
The patent transitions from one-dimensional acoustic analysis to two-dimensional analysis by incorporating visual image data. The image processor captures spatial and temporal information from video frames, adding a new dimension of information that helps distinguish true cough sounds from voiceless noise, thereby improving measurement precision
3Loss of information
If acoustic data from microphone array is used, then sound arrival direction can be estimated, but device complexity increases
Solution Approach 1:
The patent divides the detection system into separate functional modules: the acoustic feature extractor handles sound analysis, the direction estimator processes spatial information from the microphone array, and the image processor handles visual data. This segmentation allows the system to retain valuable spatial information while managing complexity through modular design
Data Source
AI summary
A cough detection device including: an acoustic feature extractor that extracts at least one acoustic feature from acoustic data output by a microphone array according to a sound received; a first identifier that performs identification of the sound based on the at least one acoustic feature to determine whether the sound is a cough sound; a direction estimator that estimates an arrival direction of the sound from the acoustic data; an image selector that selects, from first image data indicating an image obtained by capturing a scene in which the sound occurs, second image data indicating an area corresponding to the arrival direction estimated; and a second identifier that performs identification of the image based on the second image data to determine whether a coughing action is shown in the image.