ArticlePublisher preview available

Edge Computing System for Automatic Detection of Chronic Respiratory Diseases Using Audio Analysis

Authors:
José Antonio Rivas-Navarrete
José Antonio Rivas-Navarrete
José Antonio Rivas-Navarrete
  • This person is not on ResearchGate, or hasn't claimed this research yet.
Humberto Pérez Espinosa at National Institute of Astrophysics, Optics and Electronics
Humberto Pérez Espinosa
A. L. Padilla-Ortiz at National Autonomous University of Mexico
A. L. Padilla-Ortiz
Ansel Y. Rodríguez González at Center for Scientific Research and Higher Education at Ensenada
Ansel Y. Rodríguez González
Show all 5 authorsHide
Read publisher preview
Download citation
To read the full-text of this research, you can request a copy directly from the authors.

Abstract and Figures

Chronic respiratory diseases affect people worldwide, but conventional diagnostic methods may not be accessible in remote locations far from population centers. Sounds from the human respiratory system have displayed potential in autonomously detecting these diseases using artificial intelligence (AI). This article outlines the development of an audio-based edge computing system that automatically detects chronic respiratory diseases (CRDs). The system utilizes machine learning (ML) techniques to analyze audio recordings of respiratory sounds (cough and breath) and classify the presence or absence of these diseases, using features such as Mel frequency cepstral coefficients (MFCC) and chromatic attributes (chromagram) to capture the relevant acoustic features of breath sounds. The system was trained and tested using a dataset of respiratory sounds collected from 86 individuals. Among them, 53 had chronic respiratory conditions, including asthma and chronic obstructive pulmonary disease (COPD), while the remaining 33 were healthy. The system’s final evaluation was conducted with a group of 13 patients and 22 healthy individuals. Our approach demonstrated high sensitivity and specificity in the classification of sounds on edge devices, including smartphone and Raspberry Pi. Our best results for CRDs reached a sensitivity of 90.0%, a specificity of 93.55%, and a balanced accuracy of 91.75% for accurately identifying individuals with both healthy and diseased. These results showcase the potential of edge computing and machine learning systems in respiratory disease detection. We believe this system can contribute to developing efficient and cost-effective screening tools.
Figures - available from: Journal of Medical Systems
This content is subject to copyright. Terms and conditions apply.
A preview of this full-text is provided by Springer Nature.
Learn more
Content available from Journal of Medical Systems
This content is subject to copyright. Terms and conditions apply.
Journal of Medical Systems (2025) 49:33
https://doi.org/10.1007/s10916-025-02154-7
RESEARCH
Edge Computing System for Automatic Detection of Chronic
Respiratory Diseases Using Audio Analysis
José Antonio Rivas-Navarrete1·Humberto Pérez-Espinosa2·A.L. Padilla-Ortiz3,4 ·Ansel Y. Rodríguez-González1·
Diana Cristina García-Cambero5
Received: 28 February 2024 / Accepted: 27 January 2025
© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2025
Abstract
Chronic respiratory diseases affect people worldwide, but conventional diagnostic methods may not be accessible in remote
locations far from population centers. Sounds from the human respiratory system have displayed potential in autonomously
detecting these diseases using artificial intelligence (AI). This article outlines the development of an audio-based edge
computing system that automatically detects chronic respiratory diseases (CRDs). The system utilizes machine learning
(ML) techniques to analyze audio recordings of respiratory sounds (cough and breath) and classify the presence or absence of
these diseases, using features such as Mel frequency cepstral coefficients (MFCC) and chromatic attributes (chromagram) to
capture the relevant acoustic features of breath sounds. The system was trained and tested using a dataset of respiratory sounds
collected from 86 individuals. Among them, 53 had chronic respiratory conditions, including asthma and chronic obstructive
pulmonary disease (COPD), while the remaining 33 were healthy. The system’s final evaluation was conducted with a group
of 13 patients and 22 healthy individuals. Our approach demonstrated high sensitivity and specificity in the classification
of sounds on edge devices, including smartphone and Raspberry Pi. Our best results for CRDs reached a sensitivity of
90.0%, a specificity of 93.55%, and a balanced accuracy of 91.75% for accurately identifying individuals with both healthy
and diseased. These results showcase the potential of edge computing and machine learning systems in respiratory disease
detection. We believe this system can contribute to developing efficient and cost-effective screening tools.
Keywords CDR ·COPD ·Machine learning ·Edge computing
Introduction
Respiratory diseases such as asthma, chronic obstructive
pulmonary disease (COPD), and pneumonia not only sig-
nificantly affect the quality of life of those affected, but also
pose substantial economic and operational pressure on public
healthcare systems around the world [1]. Chronic respira-
tory diseases pose a significant health challenge that affects
people of all ages. According to the World Health Orga-
nization (WHO), chronic respiratory diseases caused 4.1
million deaths worldwide in one year (2019) [2]. Early detec-
tion is essential to prevent complications and improve the
prognosis of patients with these diseases. Emerging digital
tools and technologies offer the potential to improve early
diagnosis. Currently, healthcare professionals use conven-
tional diagnostic techniques, such as pulmonary function
tests, chest radiographs, CT scans, and arterial blood gas
Extended author information available on the last page of the article
analysis [3]. However, these methods can be expensive and
time-consuming, require specialized equipment and trained
professionals, and limit access to appropriate care, especially
in contexts characterized by resource scarcity.
Symptoms of chronic respiratory diseases include short-
ness of breath, wheezing, and coughing [4], which can cause
severe complications if left untreated. A lesser-known cause
of COPD is the use of firewood as cooking fuel, a common
practice in many communities, particularly in rural areas
and developing countries [5]. Current patient monitoring
approaches include self-reports conducted in a home envi-
ronment [6] and clinical examinations, often supplemented
by clinical evaluation tests [7]. However, self-reports are
inherently subjective and susceptible to data omissions, given
the inherent challenges of data collection. However, clinical
examinations can be more invasive and require expensive
equipment. In addition, clinical trials can produce inconsis-
tent results due to various factors, such as the complexity
0123456789().: V,-vol 123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
ResearchGate has not been able to resolve any citations for this publication.
Multi Feature fusion for COPD Classification using Deep learning algorithms
Article
Full-text available
  • Jan 2024
Machine learning (ML) and deep learning (DL) are becoming pivotal for providing solutions to healthcare issues. Due to their accurate and quick forecasting models and discoveries, ML and DL algorithms are being used for disease classification by healthcare experts. Along with life-threatening illnesses like cancer, respiratory problems such as Chronic Obstructive Pulmonary Disease (COPD) have been growing more prevalent and endangering the survival of human society. According to the World Health Organization, COPD will be the third-leading cause of death and the seventh-leading cause of illness globally by 2030. Therefore, early detection and fast treatment are essential. The primary methods for diagnosing COPD need inadequate and pricy spirometer and imaging equipment. In this paper, an attempt is made to determine the severity of COPD disease using ML and DL algorithms using the cough sound of the patient. To extract audio features like Mfcc, Chroma, Contract, Mel, and Tonnetz, we have used the Librosa Python Library. To address the issues of imbalanced dataset, we have used the SMOTE algorithm. To find the most effective multi feature fusion for classifying COPD, numerous experiments have been carried out using various fusions of audio features. For the purpose of evaluating the multi-feature fusion's performance, we have run MLP, CNN, RNN, and LSTM models on fusion of two audio features and three audio features. Results of experiments suggest that the LSTM model with Adam as an optimization function gives 100% training accuracy and 87% testing accuracy for fusion of Mfcc and Mel features. As a result of the fusion of the three features of Tonnetz, Chroma, and Mel, CNN model performs better with training accuracy of 90% and testing accuracy of 82%. URN:NBN:sciencein.jist.2024.v12.780
View
Bird Species Identification from Audio Data
Conference Paper
Full-text available
  • Jul 2023
Birds are an important indicator species for environmental changes, and identifying bird species can provide valuable insights into changes in their populations and habitats. This research focuses on identifying bird species using audio recordings from the BirdCLEF dataset on Kaggle. Using these audio recordings, we extracted 26 acoustic features like MFCC, spectral centroid, spectral bandwidth, etc. These features were then used to train supervised machine learning models like Decision trees, Random forests, Naive Bayes classifier, Support Vector Classifier (SVC), k-nearest neighbor (K-NN), and Stochastic Gradient Descent (SGD). In the end, we trained the models using the combined features from the original audio and cleaned audio files along with feature selection by performing recursive feature elimination (RFE).
View
View
Automated Discrimination of Cough in Audio Recordings: A Scoping Review
Article
Full-text available
  • Jun 2022
The COVID-19 virus has irrevocably changed the world since 2020, and its incredible infectivity and severity have sent a majority of countries into lockdown. The virus’s incubation period can reach up to 14 days, enabling asymptomatic hosts to transmit the virus to many others in that period without realizing it, thus making containment difficult. Without actively getting tested each day, which is logistically improbable, it would be very difficult for one to know if they had the virus during the incubation period. The objective of this paper’s systematic review is to compile the different tools used to identify coughs and ascertain how artificial intelligence may be used to discriminate a cough from another type of cough. A systematic search was performed on Google Scholar, PubMed, and MIT library search engines to identify papers relevant to cough detection, discrimination, and epidemiology. A total of 204 papers have been compiled and reviewed and two datasets have been discussed. Cough recording datasets such as the ESC-50 and the FSDKaggle 2018 and 2019 datasets can be used for neural networking and identifying coughs. For cough discrimination techniques, neural networks such as k-NN, Feed Forward Neural Network, and Random Forests are used, as well as Support Vector Machine and naive Bayesian classifiers. Some methods propose hybrids. While there are many proposed ideas for cough discrimination, the method best suited for detecting COVID-19 coughs within this urgent time frame is not known. The main contribution of this review is to compile information on what has been researched on machine learning algorithms and its effectiveness in diagnosing COVID-19, as well as highlight the areas of debate and future areas for research. This review will aid future researchers in taking the best course of action for building a machine learning algorithm to discriminate COVID-19 related coughs with great accuracy and accessibility.
View
A study of using cough sounds and deep neural networks for the early detection of Covid-19
Article
Full-text available
  • Jan 2022
The current clinical diagnosis of COVID-19 requires person-to-person contact, needs variable time to produce results, and is expensive. It is even inaccessible to the general population in some developing countries due to insufficient healthcare facilities. Hence, a low-cost, quick, and easily accessible solution for COVID-19 diagnosis is vital. This paper presents a study that involves developing an algorithm for automated and noninvasive diagnosis of COVID-19 using cough sound samples and a deep neural network. The cough sounds provide essential information about the behavior of glottis under different respiratory pathological conditions. Hence, the characteristics of cough sounds can identify respiratory diseases like COVID-19. The proposed algorithm consists of three main steps (a) extraction of acoustic features from the cough sound samples, (b) formation of a feature vector, and (c) classification of the cough sound samples using a deep neural network. The output from the proposed system provides a COVID-19 likelihood diagnosis. In this work, we consider three acoustic feature vectors, namely (a) time-domain, (b) frequency-domain, and (c) mixed-domain (i.e., a combination of features in both time-domain and frequency-domain). The performance of the proposed algorithm is evaluated using cough sound samples collected from healthy and COVID-19 patients. The results show that the proposed algorithm automatically detects COVID-19 cough sound samples with an overall accuracy of 89.2%, 97.5%, and 93.8% using time-domain, frequency-domain, and mixed-domain feature vectors, respectively. The proposed algorithm, coupled with its high accuracy, demonstrates that it can be used for quick identification or early screening of COVID-19. We also compare our results with that of some state-of-the-art works.
View
View
View
Cough Detection Using Mobile Phone Accelerometer and Machine Learning Techniques
Chapter
  • Oct 2022
Timely cough diagnosis is important due to prevalence of severe upper respiratory tract infection during COVID-19 pandemic. Cough monitoring can aid in the detection of onset of severe infections. With a variety of embedded sensors, increased memory, higher processing capabilities, and widespread availability; smartphones can be employed to detect coughs by actively monitoring the movements of chest muscles. This chapter describes a three-axis accelerometer-based approach for detecting and classifying cough occurrences in human subjects through application of machine learning algorithms to movement data. In an unconstrained environment, movement data is captured in real-time from the embedded accelerometer of a smartphone carried by subject in upper-left pocket of shirt. Preprocessing of data is done with noise reduction, standardization, and data re-sampling techniques to ensure efficient model training. Supervised machine learning approaches are utilized to classify cough into three categories of mild, chronic and breathing episodes depending upon intensity of movement in chest muscles. According to the collected data and comparative examination of training models, Random Forest surpassed other algorithms with an accuracy of 89.1% and high precision corresponding to 19,000 samples of training data. Proposed system classifies activities from movement data stored in the cloud instead of performing this task within mobile phone application to accommodate a host of users using smartphones with varying software and hardware specifications for assisted living applications.
View
Deep Dense Autoencoder Using Modulation Spectrogram for Machine Unsupervised Anomaly Detection
Chapter
  • Jan 2023
The purpose of this paper is to design an abnormal sound detection system to detect abnormal sound during mechanical operation. The system uses the modulation spectrogram as a feature of the sound signal to train a dense autoencoder. Using the development set provided by DCASE2021 to compare 7 machine types with its baseline system, the result is better than the baseline system. Among the 7 machine types, the effects of Fan, Gearbox, Pump, Slider and Valve are significantly better than the baseline system provided by DCASE2021. By comparison, we believe that the method of feature extraction has an impact on the training of the neural network. In addition, the number of layers of the neural network should not be too large.KeywordsModulation spectrogramAutoencoderUnsupervisedAnomaly detection
View
Towards using cough for respiratory disease diagnosis by leveraging Artificial Intelligence: A survey
Article
  • Feb 2022
Cough acoustics contain multitudes of vital information about pathomorphological alterations in the respiratory system. Reliable and accurate detection of cough events by investigating the underlying cough latent features and disease diagnosis can play an indispensable role in revitalizing the healthcare practices. The recent application of Artificial Intelligence (AI) and advances of ubiquitous computing for respiratory disease prediction has created an auspicious trend and myriad of future possibilities in the medical domain. In particular, there is an expeditiously emerging trend of Machine learning (ML) and Deep Learning (DL)-based diagnostic algorithms exploiting cough signatures. The enormous body of literature on cough-based AI algorithms demonstrate that these models can play a significant role for detecting the onset of a specific respiratory disease. However, it is pertinent to collect the information from all relevant studies in an exhaustive manner for the medical experts and AI scientists to analyze the decisive role of AI/ML. This survey offers a comprehensive overview of the cough data-driven ML/DL detection and preliminary diagnosis frameworks, along with a detailed list of significant features. We investigate the mechanism that causes cough and the latent cough features of the respiratory modalities. We also analyze the customized cough monitoring application, and their AI- powered recognition algorithms. Challenges and prospective future research directions to develop practical, robust, and ubiquitous solutions are also discussed in detail.
View