Automatic discrimination between cough and non-cough accelerometry signal artefacts. (July 2019)

• Record Type: Journal Article
• Authors: Mohammadi, Helia; Samadani, Ali-Akbar; Steele, Catriona; Chau, Tom
• Is Part Of: Biomedical signal processing and control. Volume 52(2019)
• Page Start: 394
• Page End: 402
• Publication Date: 2019-07
• Subjects: Accelerometry signal -- Artificial neural networks -- Cough -- Elastic net -- Binary genetic algorithm -- Principal component analysis -- Support vector machines; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2018.10.013
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Accelerometry signals were used to represent coughs, rest and non-cough artefacts. Voluntary coughs were discriminated from non-cough artefacts at 90% accuracy. Time domain features were discriminatory for voluntary cough detection. Involuntary coughs were discriminated from non-coughs at 80% accuracy. Information theoretic features were most useful for detecting involuntary coughs. Accelerometry holds promise as tool for automatic cough detection. Abstract: Cough is the forceful and rapid expulsion of air that clears the airway of foreign material, fluid or mucus, and may be symptomatic of respiratory conditions and swallowing difficulties. The evaluation of cough severity can thus inform appropriate treatment of the underlying issue, but assessment has historically been subjective (e.g., self-report) and episodic. Automatic cough detection has been proposed as a tool for long-term (e.g., nocturnal) monitoring of cough intensity and frequency, but the rejection of non-cough activities remains an elusive challenge. Cervical accelerometry is a novel tool, which measures oropharyngeal and laryngopharyngeal vibrations associated with swallowing, coughing, tongue movements and speech. We propose an automatic cough detection system that discriminates accelerometry signals associated with coughs from those representing swallows, tongue movements and speech. We consider both voluntary and reflexive coughs. Accelerometry signals were represented in term of time-frequency meta features. Using the binary genetic feature selection algorithm and a support vector machine classifier, the proposed system achieved a cough detection accuracy of 99.26 ± 0.12% when discriminating between voluntary cough and rest accelerometry signals. An accuracy of 90 ± 13.9% was achieved using elastic net feature selection and support vector machine when classifying involuntary coughs and rest signals. When discriminating between voluntary and involuntary cough versus all other non-cough artefacts, the proposed system achieved accuracies of 90.2 ± 3.6% and 80.3 ± 10.5%, respectively. Compared to current cough monitoring systems which require combinations of microphones, accelerometers, and video recorders, the proposed method requires only a single accelerometer.