Face mask effects on speaker verification performance in the presence of noise.

Due to its spread via physical contact and the regulations on wearing face masks, COVID-19 has resulted in tough challenges for speaker recognition. Masks may aid in preventing COVID-19 transmission, although the implications of the mask on system performance in a clean environment and with varying levels of background noise are unclear. The face mask has an impact on speech output. The task of comprehending speech while wearing a face mask is made more difficult by the mask's frequency response and radiation qualities, which is vary depending on the material and design of the mask. In this study, we recorded speech while wearing a face mask to see how different masks affected a state-of-the-art text-independent speaker verification system using an i-vector speaker identification system. This research investigates the influence of facial coverings on speaker verification. To address this, we investigated the effect of fabric masks on speaker identification in a cafeteria setting. These results present preliminary speaker recognition rates as well as mask verification trials. The result shows that masks had little to no effect in low background noise, with an EER of 2.4-2.5% in 20 dB SNR for both masks compared to no mask at the same level. In noisy conditions, accuracy was 12.7-13.0% lowers than without a mask with a 5 dB SNR, indicating that while different masks perform similarly in low background noise levels, they become more noticeable in high noise levels.

Preliminary Diagnosis of COVID-19 using Speech Processing Techniques

This project aims to devise an alternative for Coronavirus detection using various audio signals. The aim is to create a machine-learning model assisted by speech processing techniques that can be trained to distinguish symptomatic and asymptomatic Coronavirus cases. Here the features exclusive to the vocal cord of a person is used for covid detection. The procedure is to train the classifier using a data set containing data of people of various ages both infected and disease-free, including patients with comorbidities. We presented a machine learning-based Coronavirus classifier model that can separate Coronavirus positive or negative patients from cough, breathing, and speech recordings. The model was trained and evaluated using several machine learning classifiers such as Random Forest Classifier, Logistic Regression (LR), Decision Tree Classifier, k-nearest Neighbour (KNN), Naive Bayes Classifier, Linear Discriminant Analysis, and a neural network. This project helps track COVID-19 patients at a low cost using a non-contactable procedure and reduces the workload on testing centers. © 2023 IEEE.

BERIS: An mBERT-based Emotion Recognition Algorithm from Indian Speech

Emotions, the building blocks of the human intellect, play a vital role in Artificial Intelligence (AI). For a robust AI-based machine, it is important that the machine understands human emotions. COVID-19 has introduced the world to no-touch intelligent systems. With an influx of users, it is critical to create devices that can communicate in a local dialect. A multilingual system is required in countries like India, which has a large population and a diverse range of languages. Given the importance of multilingual emotion recognition, this research introduces BERIS, an Indian language emotion detection system. From the Indian sound recording, BERIS estimates both acoustic and textual characteristics. To extract the textual features, we used Multilingual Bidirectional Encoder Representations from Transformers. For acoustics, BERIS computes the Mel Frequency Cepstral Coefficients and Linear Prediction coefficients, and Pitch. The features extracted are merged in a linear array. Since the dialogues are of varied lengths, the data are normalized to have arrays of equal length. Finally, we split the data into training and validated set to construct a predictive model. The model can predict emotions from the new input. On all the datasets presented, quantitative and qualitative evaluations show that the proposed algorithm outperforms state-of-the-art approaches. © 2022 Association for Computing Machinery.

An Advanced Deep Learning Medical Image Recognition and Diagnosis of Respiratory System Viruses

Respiratory infections are a confusing and time-consuming task that caused recently a pandemic that affected the whole world. One of the pandemics was COVID-19 that has exposed the vulnerability of medical services across the world, particularly in underdeveloped nations. There comes a strong demand for developing new computer-assisted diagnosis tools to present cost-effective and rapid screening in locations wherein enormous traditional testing is impossible. Medical imaging becomes critical for diagnosing disease, X-rays and computed tomography (CT) scan are employed in the deep network which will be helpful in diagnosing diseases. This paper proposes a scanning model based on using a Mel Frequency Cepstral Coefficients (MFCC) features extracted from a respiratory virus CT-Scan image and then filtered by applying Gabor filter (GF). The filtered image is passed to Convolutional Neural Network (CNN) for classifying the image for the presence of a respiratory virus such as Covid, Viral Pneumonia or being a healthy normal image. The proposed system achieved a validation accuracy of 100% with an overall accuracy of 99.44%. © 2022 IEEE.

Automatic illness prediction system through speech.

Due to the COVID-19 epidemic and the curfew caused by it, many people have sought to find an ADPS on the internet in the last few years. This hints to a new age of medical treatment, all the more so if the number of internet users continues to expand. As a result, automatic illness prediction online applications have attracted the interest of a large number of researchers worldwide. This work aims to develop and implement an automated illness prediction system based on speech. The system will be designed to forecast the sort of ailment a patient is suffering from based on his voice, but this was not feasible during the trial, therefore the diseases were divided into three categories (painful, light pain and psychological pain), and then the diagnose process were implemented accordingly. The medical dataset named "speech, transcription, and intent" served as the baseline for this study. The smoothness, MFCC, and SCV properties were used in this work, which demonstrated their high representation to human being medical situations. The noise reduction forward-backward filter was used to eliminate noise from wave files captured online in order to account for the high level of noise seen in the deployed dataset. For this study, a hybrid feature selection method was created and built that combined the output of a genetic algorithm (GA) with the inputs of a NN algorithm. Classification was performed using SVM, neural network, and GMM. The greatest results obtained were 94.55% illness classification accuracy in terms of SVM. The results showed that diagnosing illness through speech is a difficult process, especially when diagnosing each type of illness separately, but when grouping the different illness types into groups, depending on the amount of pain and the psychological situation of the patient, the results were much higher.

Domestic Violence Detection Using Smart Microphones

Domestic violence between partners and family members is a worldwide problem increasing every day. As per academic studies and media articles, it escalated during the COVID-19 outbreak. Domestic violence can portray verbally and physically in several ways (for instance, between partners or against children and older people). Deep Learning (DL) combined with the Internet of Things (IoT) technology could support the detection of domestic violence, which is one of many societal issues. This paper describes a system that uses a Deep Learning model and smart microphones to identify domestic violence. The datasets used are from the Google AudioSet (GA) and from the Toronto Emotional Speech Set (TESS). For the training of the dataset, the system used spectrograms and MFCCs (Mel-Frequency Cepstral Coefficients). The system employs two approaches: (i) an ANN (Artificial Neural Network) model, and (ii) a ResNet model. The Resnet model obtained an accuracy of 71%. The ANN model, which brought an accuracy of 83%, was tested and loaded to a Raspberry Pi, i.e., connected to the microphone for audio recording. The recorded audio was fed to the trained model, classifying the audio, and alerting the domestic violence to relatives, friends, or volunteers registered with the system via e-mail. The designed system is compact and can be placed inside the home. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Artificial Intelligence Based CoughMonitoring System For Covid-19

Infection with the SARS-CoV-2 virus results in Covid 19, an infectious illness. Most persons who get Coronavirus will only experience mild to moderate symptoms and will get better without any special care. Some people get very sick and need medical attention. The rising mortality toll from COVID-19 underscores the importance of developing methods for early detection of the disease, which might aid in containing the epidemic and facilitating the creation of tailored mitigation strategies. Current research in chaotic dynamics indicates that coughs and other vocal sounds include lung health data that can be used for symptomatic reasons. Mel frequencies Cepstral Coefficients (MFCC) are applied to cough samples, and then the audio data from coughs is fed into a GridsearchCV model using a KNN-based classification method. Our model was developed using 217 samples from training data and 55 from testing data. Cough tests conducted on both males and females are included in the dataset. An evaluation found that the model had an accuracy of 83.3%. © 2022 IEEE.

A Novel Snore Detection and Suppression Method for a Flexible Patch With MEMS Microphone and Accelerometer

Sleep apnea impacts more and more people all over the world, and obstructive sleep apnea of which is the most frequent. Hence, research on snoring detection and related suppression methods is extremely urgent. In this article, a novel low-cost flexible patch with MEMS microphone and accelerometer is developed to detect snore event and sleeping posture, and a small vibration motor embedded in the patch is designed to suppress snoring. Theoretical analyses of short-time energy, piecewise average filtering (PAF), and Mel-frequency cepstral coefficients (MFCCs) processing are described in detail, and the improved MFCCs are put forward and used as the input of the convolutional neural network (CNN). Furthermore, the snore recognition method based on the combination of similarity analysis and CNN analysis is presented, followed by the snoring suppression method. Experimental results demonstrate that the main features of the sound signals can be extracted effectively by PAF and MFCCs processing, and the data compression ratio is about 99.41%. Besides, the locations of the eigenvectors can be found accurately based on short-time energy analysis. The numbers of high similarity of snoring signals within 30 s are larger than 3, while those of non-snoring signals are often less than 3. If the preliminary screening with similarity analysis is passed, CNN analysis will be conducted to judge whether there are snoring events. The accuracy of snore recognition with CNN analysis is calculated to be as high as 99.25%. Finally, the average snoring time measured by the smart patch with snoring suppression is reduced to 15 from 135 min, which indicates that the proposed snore recognition and suppression methods are effective.

MFCC-based deep convolutional neural network for audio depression recognition

Emotions are a key factor affecting online learning, which has become the new normal due to the COVID-19 pandemic. Identifying students' emotions has important educational implications. Long-term negative emotions can lead to depression, and early identification can help reduce stress and improve learning efficiency. For depressed students in desperate need, current diagnoses are expensive and highly subjective. In response to the above problems, this paper proposes a method for automatic diagnosis of depression with the help of deep convolutional neural networks (Convolutional Neural Network, CNNs). This method inputs the Mel-frequency cepstral coefficients (MFCC) feature maps generated by the preprocessed speech into residual CNNs for training, and adjusts the number of layers and nodes of the network to achieve the highest recognition accuracy. The experimental results are achieved through different residual depths. In ResNet-34, the highest accuracy rate can reach 77%. This will effectively promote the process of emotion recognition research and promote the efficient development of online learning. © 2022 IEEE.

Domestic Violence Detection Using Smart Microphones

Domestic violence between partners and family members is a worldwide problem increasing every day. As per academic studies and media articles, it escalated during the COVID-19 outbreak. Domestic violence can portray verbally and physically in several ways (for instance, between partners or against children and older people). Deep Learning (DL) combined with the Internet of Things (IoT) technology could support the detection of domestic violence, which is one of many societal issues. This paper describes a system that uses a Deep Learning model and smart microphones to identify domestic violence. The datasets used are from the Google AudioSet (GA) and from the Toronto Emotional Speech Set (TESS). For the training of the dataset, the system used spectrograms and MFCCs (Mel-Frequency Cepstral Coefficients). The system employs two approaches: (i) an ANN (Artificial Neural Network) model, and (ii) a ResNet model. The Resnet model obtained an accuracy of 71%. The ANN model, which brought an accuracy of 83%, was tested and loaded to a Raspberry Pi, i.e., connected to the microphone for audio recording. The recorded audio was fed to the trained model, classifying the audio, and alerting the domestic violence to relatives, friends, or volunteers registered with the system via e-mail. The designed system is compact and can be placed inside the home. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Diagnosis of Covid-19 by Sound Analysis

COVID-19 virus;has dragged the world into an epidemic that has infected more than 413 million people and caused the death of nearly 6 million people. Although biomedical tests provide the diagnosis of COVID-19 with high accuracy in the diagnosis of the disease, it increases the risk of infection due to the fact that it is a method that requires contact. Machine learning models have been proposed as an alternative to biomedical testing. Cough has been identified by the World Health Organization as one of the symptoms of COVID-19 disease. In this study, the success performance of the positive case situation with machine learning was examined using the COUGHVID dataset with cough voice recordings. In order to increase the performance of the model, MFCC, Δ-MFCC and Mel Coefficients attributes were obtained after preprocessing the sound recordings. In the ensemble learning model, features were used as independent variables and a value of 0.65 AUC-ROC was reached. In addition to these performance-enhancing changes, since the acoustic properties of male and female cough sounds are different, the training of persons was carried out separately from each other, and AUC-ROC values of 0.70 for females and 0.68 for males were obtained. Trimming the silent regions at the beginning and end of the recordings, using the ensemble learning model, and grouping based on gender provided better results for this study compared to previous studies. © 2022 IEEE.

Deep learning and machine learning-based voice analysis for the detection of COVID-19: A proposal and comparison of architectures.

Alongside the currently used nasal swab testing, the COVID-19 pandemic situation would gain noticeable advantages from low-cost tests that are available at any-time, anywhere, at a large-scale, and with real time answers. A novel approach for COVID-19 assessment is adopted here, discriminating negative subjects versus positive or recovered subjects. The scope is to identify potential discriminating features, highlight mid and short-term effects of COVID on the voice and compare two custom algorithms. A pool of 310 subjects took part in the study; recordings were collected in a low-noise, controlled setting employing three different vocal tasks. Binary classifications followed, using two different custom algorithms. The first was based on the coupling of boosting and bagging, with an AdaBoost classifier using Random Forest learners. A feature selection process was employed for the training, identifying a subset of features acting as clinically relevant biomarkers. The other approach was centered on two custom CNN architectures applied to mel-Spectrograms, with a custom knowledge-based data augmentation. Performances, evaluated on an independent test set, were comparable: Adaboost and CNN differentiated COVID-19 positive from negative with accuracies of 100% and 95% respectively, and recovered from negative individuals with accuracies of 86.1% and 75% respectively. This study highlights the possibility to identify COVID-19 positive subjects, foreseeing a tool for on-site screening, while also considering recovered subjects and the effects of COVID-19 on the voice. The two proposed novel architectures allow for the identification of biomarkers and demonstrate the ongoing relevance of traditional ML versus deep learning in speech analysis.

Advanced Medical Images Recognition and Diagnosis of Respiratory System Viruses

Detection of respiratory viruses is a perplexing task which regularly requires saving time by taking a quick look at clinical images of patients ceaselessly. Hence, there's a need to propose and develop a model to predict the respiratory viruses (COVID-19) cases at the earliest possible to control the spread of disease. Deep learning makes it possible to find out that Covid-19 can be detected in an efficient way using its classification tools such as CNN (Convolutional Neural Network). MFCC (Mel Frequency Cepstral Coefficients) is a very common and efficient technique for signal processing. In this research, a MFCC - CNN learning model to hasten the prediction process is proposed that assist the medical professionals. MFCC is used for extracting the image's features concerning existence of COVID-19 or not. Classification is performed by using convolutional neural network. This makes the time-consuming process easier and faster with more accurate results for radiologists and this reduces the spread of virus and save lives. Experimental results shows that using CT image converted to Mel-frequency cepstral coefficient spectrogram images as input to a CNN can achieve a high accuracy results;with classification of validation data scoring an accuracy of 99.08% correct classification of COVID and NON COVID labeled images. Hence, it can be used practically for detection of COVID-19 from CT images. The work here provides a proof of concept that high accuracy can be achieved with a moderate dataset, which can have a significant impact in this area. © 2022 Croatian Society MIPRO.

Covid Cough Classification using KNN Classification Algorithm

Covid 19 is an infectious disease that is caused by infection due to SARS-CoV-2 virus. The vast majority of people infected with Corona virus will encounter mild to moderate symptoms and recover without any special treatment. In some case, some people become seriously ill and require clinical consideration. Because of the increase in number of death due to COVID-19, an techniques for the early discovery of the illness is very much needed that might assist with restricting its spread just as help in the development of targeted surrounding solutions. Coughs and other vocal sounds contain pulmonary health data that can be utilized for symptomatic purposes, and ongoing examinations in chaotic dynamics have shows a nonlinear phenomenon exists in vocal signs. Cough samples are transformed with Mel frequency Cepstral Coefficients (MFCC) and the cough audio data is fitted into a GridsearchCV model with KNN based classification algorithm. The number of training data for used for training our model is 217 and remaining 55 data were used for testing the model. The dataset contains the cough tests from both male and female. When evaluated the model could get a precision of 83.3%. © 2022 IEEE.

Automated Covid-19 Respiratory Symptoms Analysis from Speech and Cough

An effective contact tracking strategy helps to maintain control over the Covid-19 pandemic. People without visible symptoms make it a complex problem because there has to be an unobtrusive way to discover that they are virus carriers and have to be isolated. Automated Covid-19 respiratory symptoms analysis helps to focus on people with respiratory symptoms. In our approach, a telephone call system leads a dialog and discovers Covid-19 disease by analyzing a person's speech and cough. After a positive match, it invites the person to PCR testing to confirm or reject the diagnosis. We compare our speech and cough detection system with Interspeech Computational Paralinguistics Challenge (ComParE) 2021. The results of Covid-19 Speech Sub-Challenge sub-challenge (CSS) show that we outperform the baseline results by 4.3% of the Unweighted Average Recall value. © 2022 IEEE.

Accent Classification of Native and Non-Native Children using Harmonic Pitch

To combat the Covid-19 outbreak, the education system shifted away from the classroom to distinct e-learning on digital platforms, which made effective use of voice-based recognition systems, especially for preliterate children. Children’s speech recognition systems face multiple challenges owing to their immature vocal tracts, and they demand more intelligence due to the fact that children with diverse accents utter words differently. Accent refers to a unique style of pronouncing a language, particularly one associated with a specific nation, place, or socio-economic background. This paper aims to extract reliable acoustic and prosodic speech cues of accent for classification of native and non-native preschool children using harmonic pitch estimation along with Mel Frequency Cepstral Coefficients (MFCCs) to train the k-Nearest Neighbour (k-NN) classifier. The experimental results reveal that the proposed robust model outperforms various feature extractors in accent classification of native and non-native children in terms of accuracy & F-Measure and more discriminate against noisy environments. © 2022 IEEE.

Covid-19 Detection using Spectral and Statistical Features of Cough and Breath Sounds

The pandemic situation due to Corona Virus Disease of 2019 (COVID-19) is significant public health risk around the world. The infected people can spread this virus very quickly. Due to this reason, the early detection is essential to reduce its spread. This research effort aims to develop a method for diagnosis of COVID-19 based on the recording of cough and breath sounds. In this paper, a convolutional neural network (CNN) classifier is applied after train and test splitting for cough and breath sound features. The present work show that the combination of MFCC and cepstrum-based statistical features along with ZCR improve the accuracy of detection to the great extent. It shows great potential in the development of automatic COVID-19 detection tool.

Remote monitoring of Covid-19 patients using IoT and AI

The covid-19 pandemic has brought changes in various sectors like healthcare, business, education and economy. Due to the large spread of covid-19 in a lot of countries there is shortage of hospital beds, oxygen supply and healthcare workers. So, the pandemic generated need to use smart pioneering technologies like Artificial Intelligence and Internet of Things to monitor patient in an effective way. In this research paper a prototype is proposed based on IoT and AI for monitoring home quarantine covid-19 patients. Wearable IoT devices automatically collect information like oxygen level, temperature of body, etc. with the help of integrated sensors. Coughing is one of the most noticeable symptoms of people infected with covid-19. Frequency of cough is detected using Tensor flow library of Deep learning model. This prototype is a way to make IoT sensors smarter enough to detect coughs with the help of a trained model. Coughing dataset is collected and labelled manually. Dataset is self-created and categorized into cough and noise. Cough detection is based on MFCC features using DNN and CNN. The use of these technologies can bring a quick transition in healthcare to avoid risks caused with the life of human beings. © 2022 IEEE.

Human-Robot Interaction in Bengali language for Healthcare Automation integrated with Speaker Recognition and Artificial Conversational Entity

The research study presents an architecture of HumanRobot Interaction (HRI) based Artificial Conversational Entity integrated with speaker recognition ability to avail modern healthcare services. Due to the Covid-19 pandemic, the situation has become troublesome for health workers and patients to visit hospitals because of the high risk of virus dissemination. To minimize the mass congestion, our developed architecture would be an appropriate, cost-effective solution that automates the reception system by enabling AI-based HRI and providing fast and advanced healthcare services in the context of Bangladesh. The architecture consists of two significant subsections: Speaker Recognition and Artificial Conversational Entities having Automatic Speech Recognition in Bengali, Interactive Agent, and Text-to-Speech-synthesis. We used MFCC features as the linguistic parameters and the GMM statistical model to adapt each speaker's voice and estimation and maximization algorithm to identify the speaker's identity. The developed speaker recognition module performed significantly with 94.38% average accuracy in noisy environments and 96.27% average accuracy in studio quality environments and achieved a word error rate (WER) of 42.15% from RNN based Deep Speech 2 model for Bangla Automatic Speech Recognition (ASR). Besides, Artificial Conversational Entity performs with an average accuracy of 98.58% in a small-scale real-time environment. © 2021 IEEE.

Deep Learning-based Mixed Data Approach for COVID-19 Detection

The outbreak of COVID-19 has caused an exponential increase in mortality rate globally and has dealt a devastating blow to nations all over the world. This unforeseen calamity needs to be tackled and early detection of this disease could help in this regard. Several research studies used Chest X-rays and CT scans to detect the disease, which can be made cost-effective by using cough samples. These systems can further be refined by using multiple health parameters to provide more accurate results. In this view, this paper proposes a constructive way for the early detection of COVID-19 by considering cough samples and clinical data (Saturation of Peripheral Oxygen (SpO2) level, body temperature, heart rate, and symptoms). The dataset was collected by using a Raspberry Pi and an online questionnaire. In this paper, we put forward two approaches being Manual feature extraction and Mixed data neural networks (Multi-layer Perceptron and Convolutional Neural Networks) for efficiently handling the problem. To help the user access the system more comfortably, a mobile application was developed. The Mixed data neural networks yielded the best performance with an Area Under the Curve (AUC) score of 0.94 and an accuracy of 0.85. © 2021 IEEE.