Humans can identify a speaker by listening to their voice, over the telephone, or on any digital devices. Acquiring this congenital human competency, authentication technologies based on voice biometrics, such as automatic speaker recognition (ASR), have been introduced. An ASR recognizes speakers by analyzing speech signals and characteristics extracted from speaker’s voices. ASR has recently become an effective research area as an essential aspect of voice biometrics. Specifically, this literature survey gives a concise introduction to ASR and provides an overview of the general architectures dealing with speaker recognition technologies, and upholds the past, present, and future research trends in this area. This paper briefly describes all the main aspects of ASR, such as speaker identification, verification, diarization etc. Further, the performance of current speaker recognition systems are investigated in this survey with the limitations and possible ways of improvement. Finally, a few unsolved challenges of speaker recognition are presented at the closure of this survey.

/pdf/a-survey-of-speaker-recognition-fundamental-theories-4uxjyvlahi.pdf

A Survey of Speaker Recognition: Fundamental Theories, Recognition Methods and Opportunities

From patterns to patients: Advances in clinical machine learning for cancer diagnosis, prognosis, and treatment

In breast cancer screening, binary classification of mammograms is a common task aiming to determine whether a case is malignant or benign. A Computer-Aided Diagnosis (CADx) system based on a trainable classifier requires clean data and labels coming from a confirmed diagnosis. Unfortunately, such labels are not easy to obtain in clinical practice, since the histopathological reports of biopsy may not be available alongside mammograms, while normal cases may not have an explicit follow-up confirmation. Such ambiguities result either in reducing the number of samples eligible for training or in a label uncertainty that may decrease the performances. In this work, we maximize the number of samples for training relying on multi-task learning. We design a deep-neural-network-based classifier yielding multiple outputs in one forward pass. The predicted classes include binary malignancy, cancer probability estimation, breast density, and image laterality. Since few samples have all classes available and confirmed, we propose to introduce the uncertainty related to the classes as a per-sample weight during training. Such weighting prevents updating the network's parameters when training on uncertain or missing labels. We evaluate our approach on the public INBreast and private datasets, showing statistically significant improvements compared to baseline and independent state-of-the-art approaches. Moreover, we use mammograms from Susan G. Komen Tissue Bank for fine-tuning, further demonstrating the ability to improve the performances in our multi-task learning setup from raw clinical data. We achieved the binary classification performance of AUC = 80.46 on our private dataset and AUC = 85.23 on the INBreast dataset.

https://www.frontiersin.org/articles/10.3389/fradi.2021.796078/pdf

Leveraging Multi-Task Learning to Cope With Poor and Missing Labels of Mammograms

This study proposes an approach to voice identification based on neural networks (DNN) for i-Vector. Modern voice identification systems based on DNN use large amounts of labeled training data. Using the LRE i-Vector Machine Learning Challenge restricts access to ready-to-use i-Vector for learning and testing the voice identification system. This poses unique challenges in developing DNN-based voice identification systems, since optimized external interfaces and network architectures can no longer be used. We propose to use the training i-Vectors to train the initial DNN to identify the voice. Next, we present a novel strategy for using this initial DNN to strip the language labels of the inappropriate set from the development data. The final DNN for voice identification is trained using the original training data and the estimated out-of-set language data. We show that augmenting the training set with out-of- set labels leads to a significant improvement in voice identification performance. In this paper, we studied the possibility of using neural networks for speech identification. In particular, standard approaches to speech recognition were considered, the concept of an artificial neuron as an object used in speech identification was defined. A speech recognition option using a neural network was investigated, and steps were presented to perform this task. Accuracy using neural networks with limited learning data and a higher i-vector dimension is superior to others with a score of 92.1%. From this study, we can conclude that the size of the UBM and the dimension of the i-vector affect the accuracy of voice identification based on the i-vector.

/pdf/voice-verification-using-i-vectors-and-neural-networks-with-4trl38mcmt.pdf

Voice verification using i-vectors and neural networks with limited training data

In this digital era, developing a secure as well as democratically responsible vote casting system is becoming an increasingly challenging task. This research study attempts to integrate the blockchain technology to develop a secured and distributed electronic voting system. The primary objective of this research study is to develop a decentralized voting system, which can easily support an open, rational, and easily verifiable democratic plan. The obtain results prove that the proposed smart and reliable electronic voting machine is beneficial for the current and future democratic framework by implementing an essential E-vote casting features along with a provision to allow the citizens to change/update their vote within the given time slot.

Blockchain based Reliable Electronic Voting Technology

An individual Human computer interference system is being introduced. In olden times, as an input device the mouse and keyboard were used by human computer interference system. Those people who are suffering from certain disease or illness cannot be able to operate computers. The idea of controlling the computers with the eyes will serve a great use for handicapped and disabled person. Also this type of control will eliminate the help required by other person to handle the computer. This measure will be the most useful for the person who is without hands through which they can operate with the help of their eye movements. The movement of the cursor is directly associated with the center of the pupil. Hence our first step would be detecting the center of point pupil. This process of pupil detection is implemented using the Raspberry Pi and OpenCV. The Raspberry Pi has a SD/MMC card slot which is used for placing the SD card. The SD card is boosted with the operating system that is required starting up of Raspberry Pi. The Raspberry PI will get executed once the application program is loaded into it.

Eyeball based Cursor Movement Control

The medical field is one of the most important are as that is researched by numerous researchers in order to develop the existing system. Though there have been numerous new and innovative systems developed, there are always some challenges present in the field that needs to be solved. It is important for a healthcare system to make sure that it is secure enough and any leak of the patient data could lead to numerous other risk factors. In this paper, we propose a speaker recognition system that makes use of the Gaussian Mixture Model which identifies a person from characteristics of their voice to authenticate them to access the medical records. The model is able to determine the gender of the person to be authenticated and also recognizes the person. The model uses Cepstral Analysis as the feature extraction technique using the feature Mel Frequency Cepstral Coefficient (MFCC). The proposed work is observed to work well when compared with other existing systems and could be of much use in performing and giving access to the authorized employees in a health care system.

Speaker Recognition System using Gaussian Mixture Model

Proposed approach illustrates Detection touch-less heartbeat and a cardiopulmonary sign A microwave frame is attempted to find a good way off of 1 m from the pulse sound of a person by using a vector organize analyzer The proposed method requires the ability to distinguish pulse patterns with both recurrence and the plausibleness of the force tuning Calculations are made at 24, 58, 10, 16, and 60 GHz, and also at varying force rates between 0 and −27 dBm A classification that applies cardiopulmonary activity is defined in terms of the air and heart beatings statistics Using wavelet and wide channels, heartbeat rate and heart rate changeability are isolated from the test signal for SNRs between 0 and −20 dB

Touch-Less Heart Beat Detection on Improved LBP Algorithm

Breast cancer is the most dangerous leading cause of cancer-related death among women in the world. Internet of things (IoT) connected device that can help detect breast cancer earlier. Mammograms are the main way to detect breast cancer. But it is difficult to detect the cancer cell in the breast tissue. Because it has more tissue and less fat, the digitized mammography images are used for analyzing the abnormal areas of density, mass, and calcification that indicate the presence of cancer. In cases where adequate labeled data is available, deep learning approaches show considerable progress, and many advanced deep learning approaches have been proposed that have shown efficiency in the last few years in various modalities of computer vision and medical imaging. The small size of the medical image dataset also lowers the efficiency and robustness of computer-aided detection and/or diagnosis (CAD) systems based on deep learning. This research aims to build and test a new hybrid deep learning-based CAD method to predict the probability of a breast lesion found on a mammogram being malignant in an attempt to overcome this technological challenge. In this method, a deep convolutional neural network (CNN) will first be pre-trained using the ImageNet dataset. This work aims to develop and test a new CAD method based on hybrid transfer learning for mammographic mass classification. By combining the initial mammogram images with their morphological and texture variations, it is possible to produce more pseudo-color images to be added with information. Additional filters are used to remove unrelated transferred CNN characteristics on the network.

Deep Learning Modeling Using Normal Mammograms for Predicting Breast Cancer

New opportunities for clinical data processing and accessibility for patients to view and share their health data are opening up through advances in electronic medical reports, cloud data storage, and patient data protection regulations. Securing data, processing, transfers, and maintaining their seamless integration are of tremendous importance for any data-driven enterprise. Particularly in healthcare, blockchain technology can robustly and effectively solve these essential problems. Blockchain technology has entered a big boom in the health sector amid its utility in solving the safety concerns of the EHR in healthcare. EHRs are able to improve the delivery of healthcare. The introduction of EHRs offered the potential for information technologies to counteract those efforts. Electronic access to medical records allowed doctors to increase the quality of care significantly. In fact, EHR allows improved control of the condition and increased levels of preventive treatment. The digital record provides better functions to support decision-making and enhanced collaboration among carers. Therefore, its position in the health sector is increasingly recognized. The protection of such digital information is, therefore, given the utmost importance, and blockchain is widely used for safe and stable data on healthcare. Holding data integrity and rising data processing storage costs in blockchain healthcare technology play a significant role. Encrypted in the block chain, the proposed LWKG architecture helps a customer to show the moment that health data is displayed. LWKG verifies the integrity of information and analyzes the results stored in the database.

T. Anandhi

Papers

Eyeball based Cursor Movement Control

Speaker Recognition System using Gaussian Mixture Model

Touch-Less Heart Beat Detection on Improved LBP Algorithm

Deep Learning Modeling Using Normal Mammograms for Predicting Breast Cancer

Security Framework for Enhancing Security and Privacy in Healthcare Data Using Blockchain Technology