Novel Feature Extraction Algorithm using DWT and Temporal Statistical Techniques for Word Dependent Speaker’s Recognition

Singh, Ratnesh Kumar; Saha, Rajesh; Pal, Pankaj Kumar; Singh, Ghanapriya

doi:10.1109/icrcicn.2018.8718681

Cited by 6 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…17 When computing complexity is considered, retrieved characteristics range from straightforward statistical measures, like (Standard Deviation) SD and mean to frequently modality-dependent complex features like the number of SCR peaks. 18 While in DL approaches, classification, and feature extraction are handled by the algorithm itself, eliminating the requirement for manual feature extraction. 19 There are mental healthcare strategies that incorporate stress awareness as a key component; however, these studies are catered to a certain audience.…”

Section: Related Workmentioning

confidence: 99%

“…Linear or non‐linear, time‐ or frequency‐domain, unimodal or multimodal, and other categories can be used to classify extracted properties 17 . When computing complexity is considered, retrieved characteristics range from straightforward statistical measures, like (Standard Deviation) SD and mean to frequently modality‐dependent complex features like the number of SCR peaks 18 . While in DL approaches, classification, and feature extraction are handled by the algorithm itself, eliminating the requirement for manual feature extraction 19 …”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Hybrid deep learning model for wearable sensor‐based stress recognition for Internet of Medical Things (IoMT) system

Singh,

Chetia Phukan,

Gupta

et al. 2023

Int J Communication

Self Cite

View full text Add to dashboard Cite

SummaryThere are numerous, interrelated, and multi‐dimensional aspects that influence a person's mental health, one of them being stress. Smart wearable technology having physiological and motion sensors has paved the way for real‐time data collection to deliver cutting‐edge information about the stress of individuals. It is now possible to build an Internet of Medical Things (IoMT) system that can recognise the user's stress, revealing the elements that cause stress. However, there are significant gaps in the existing system for stress recognition. To begin with, stress recognition is primarily studied for a specific set of people, such as occupational stress among office working persons or hospital staff during emergency duties, and so forth. Second, most past work on stress recognition has focused on extracting handcrafted traits, which necessitates human interaction and skill. To overcome above mentioned challenges, this work proposes a novel IoMT framework for continuous stress recognition for mental well‐being. This paper presents a hybrid deep learning (DL) approach for automatically retrieving features and classification into various stress states for the IoMT system to address these difficulties. The proposed expanded system gathers data from wearable physiological sensors and feeds it into a convolutional neural network–long short‐term memory (CNN‐LSTM), a hybrid DL classifier. The suggested approach has been tested on the wearable stress and affect detection (WESAD) dataset. It has a reported accuracy of 90.45%, which is higher than previously reported accuracies from existing machine learning (ML) and DL approaches.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Hybrid deep learning model for wearable sensor‐based stress recognition for Internet of Medical Things (IoMT) system

Singh,

Chetia Phukan,

Gupta

et al. 2023

Int J Communication

Self Cite

View full text Add to dashboard Cite

show abstract

“…Additionally, Linear Prediction Coefficients (LPC) and Linear Spectral frequencies (LSF) were used for different applications like speaker recognition [15], spoken digits recognition [16], and emotion recognition from speech [17]. Discrete Wavelet Transform is another feature extraction technique that has been used for speaker recognition [18], speech semantic and emotions recognitions [19,20]. Furthermore, spectrogram images are the best choice of speech and acoustic feature extraction that is suitable for CNN models.…”

Section: Introductionmentioning

confidence: 99%

Design and Implementation of Fast Spoken Foul Language Recognition with Different End-to-End Deep Neural Network Architectures

Wazir

Karim

Abdullah

et al. 2021

Sensors

View full text Add to dashboard Cite

Given the excessive foul language identified in audio and video files and the detrimental consequences to an individual’s character and behaviour, content censorship is crucial to filter profanities from young viewers with higher exposure to uncensored content. Although manual detection and censorship were implemented, the methods proved tedious. Inevitably, misidentifications involving foul language owing to human weariness and the low performance in human visual systems concerning long screening time occurred. As such, this paper proposed an intelligent system for foul language censorship through a mechanized and strong detection method using advanced deep Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) through Long Short-Term Memory (LSTM) cells. Data on foul language were collected, annotated, augmented, and analysed for the development and evaluation of both CNN and RNN configurations. Hence, the results indicated the feasibility of the suggested systems by reporting a high volume of curse word identifications with only 2.53% to 5.92% of False Negative Rate (FNR). The proposed system outperformed state-of-the-art pre-trained neural networks on the novel foul language dataset and proved to reduce the computational cost with minimal trainable parameters.

show abstract

“…The sensors on wearable device provide raw data on which various signal processing algorithms such as integer wavelet transform (Singh, Singh, et al, 2020b), local mean decomposition (Kaloni et al, 2021), correlation based features (Gupta et al, 2021), are derived so as to eliminate unwanted noise and extract useful information. These features are then provided to various algorithms such as machine learning classifiers (Singh et al, 2018), transfer learning (Agarwal et al, 2020), deep learning (DL) algorithms (Pandey et al, 2020), evolutionary algorithms (Singh & Rawat, 2013) and so forth that has the capability to intelligently and automatically assign test windowed time series signal into different classes.…”

mentioning

confidence: 99%

Stress recognition with multi‐modal sensing using bootstrapped ensemble deep learning model

2023

View full text Add to dashboard Cite

The factors that influence a person's mental health are numerous, interconnected, and multi-dimensional. Recognition of stress is one of the facets in developing the Mental Healthcare (MHC) system framework. With the advent of technology, smart wearable devices have paved a way to collect data in real-time to provide the cutting-edge reports about the individual. Due to the physiological sensors present in the smart wearable devices, it is now possible to have a robust system to recognize the stress of the smart wearable devices user thus consecutively leading to recognition of factors in leading to stress. However, the current MHC system for recognition and identification of stress have several drawbacks. First, stress recognition is mostly designed for a particular group of individuals like occupational stress, perinatal maternal stress, or health worker stress and fails to propose a framework that would not be targeted for a particular group of individual. Second, most of the previous work done on stress recognition focuses on the extraction of handcrafted features thus requiring human intervention and expertise. To address these issues, this study proposes, a hybrid deep learning based ensemble approach for automated extraction of features and classification into various state of stress for MHC system. The proposed framework takes input from wearable physiological sensors and is provided to deep learning classifier of convolutional neural network (CNN) and CNN-long short term memory based ensemble model. The proposed framework has been experimented on the wearable stress and affect detection dataset and reports an accuracy of 91.52% that is 7.20% higher than earlier reported accuracies from other machine learning and deep learning models.

show abstract

Novel Feature Extraction Algorithm using DWT and Temporal Statistical Techniques for Word Dependent Speaker’s Recognition

Cited by 6 publications

References 11 publications

Hybrid deep learning model for wearable sensor‐based stress recognition for Internet of Medical Things (IoMT) system

Hybrid deep learning model for wearable sensor‐based stress recognition for Internet of Medical Things (IoMT) system

Design and Implementation of Fast Spoken Foul Language Recognition with Different End-to-End Deep Neural Network Architectures

Stress recognition with multi‐modal sensing using bootstrapped ensemble deep learning model

Contact Info

Product

Resources

About