Jehosheba Margaret M

Ms. Jehosheba Margaret M

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
14 May 2023 in Brain-Computer Interfaces
Verified
Journal Article
Closed access
JMJehosheba Margaret M
Jehosheba Margaret M
NBN.M Masoodhu Banu
N.M Masoodhu Banu
Emotion is an important factor that decides the the state of the mind of an individual. However, there are many people who cannot express their emotions explicitly due to various psychological or physiological issues. The recent technology development in interdisciplinary techniques has made the emotion recognition easy. The main objective of this paper is to uncover the contribution of hybrid deep learning model in classifying the emotions. This novel hybrid Bi-LSTM model is applied to three dimensional VAD model to classify 16 emotions. In addition, this paper addresses their performance comparison with respect to other models. Deep learning models like CNN, LSTM, Hybrid Bi-LSTM, and Hybrid Bi-GRU were used for experimentation. FFT is used to convert from time domain to frequency domain in all the models. The performance measure of these models is estimated in terms of accuracy, precision, recall, and F1score. The estimated accuracy of different deep learning models are approximated to be 87.5% for CNN, 88.7 % for LSTM, 93.9% for Hybrid Bi-LSTM, and 92.2% for Hybrid GRU based on the number of subjects. These comparisons have helped to find the suitable deep learning model for emotion recognition.

31 August 2024 in Traitement du Signal
Verified
Journal Article
Open access
JMJehosheba Margaret M
Jehosheba Margaret M
MMMasoodhu Banu Noordheen Mohammad Mustafa
Masoodhu Banu Noordheen Mohammad Mustafa

21 September 2023 in International Journal on Recent and Innovation Trends in Computing and Communication
Verified
Journal Article
Open access
PDF file
SRShelishiyah R
Shelishiyah R
DTDeepa Beeta Thiyam
Deepa Beeta Thiyam
JMJehosheba Margaret M
Jehosheba Margaret M
The Brain-Computer Interface (BCI) technologies have excellent clinical and non-clinical uses. Among the most popular imaging methods adopted in BCI technologies is electroencephalography (EEG). But EEG signals are typically quite complicated, so analyzing them necessitates a significant amount of effort. With the help of machine learning (ML), this research investigates the feasibility of a BCI platform based on the motor imagery (MI) concept. The steps of pre-processing, feature extraction and classification are the underpinning of any conventional ML model. To train such a model, however, a large amount of data is needed. To address this gap, this work introduces a new mayfly-optimized multiclass weighted random forest (MFO-MWRF) technique that uses retrieved features as input to mitigate the need for this supplementary data. In this study, we gather a dataset of hybrid EEG and fNIRS motor imagery that can be pre-processed using a Wiener filter (WF) to filter out noisier signals without affecting the high-quality images. The characteristics are extracted using the discrete wavelet transform (DWT). The research results indicate that the proposed approach achieves the best performance compared to existing approaches for classifying motor movement images.

15 June 2023 in Research Square
Verified
Preprint
Open access
PDF file
JMJehosheba Margaret M
Jehosheba Margaret M
MBMasoodhu Banu
Masoodhu Banu
Deep Neural Network (DNN) is an advancing technology that improves our life by allowing machines to perform complex tasks. Hybrid Deep Neural Network (HDNN) is widely used for emotion recognition using EEG signals due to its increase in performance than DNN. Among several factors that improve the performance of the network, activation is an essential parameter that improves the model accuracy by introducing non-linearity into DNN. Activation function enables non-linear learning and solve the complexity between the input and output data. The selection of activation function depends on the type of data that is used for computation. This paper investigates the model performance of different activation functions like ReLU, ELU and tanh on a hybrid CNN with Bi-LSTM model for emotion recognition. The model was tested on DEAP dataset which is an emotion dataset that uses physiological and EEG signals. The experimental results have shown that the model has improved accuracy when ELU function is used.