Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
24 results
Search Results
Item Word boundary estimation for continuous speech using higher order statistical features(Institute of Electrical and Electronics Engineers Inc., 2017) Naganoor, V.; Jagadish, A.K.; Chemmangat, K.Detection of the start and the end time of words in a continuous speech plays a crucial role in enhancing the accuracy of Automatic Speech Recognition (ASR). Hence, addressing the problem of efficiently demarcating word boundaries is of prime importance. In this paper, we introduce two new acoustic features based on higher order statistics called Density of Voicing (DoV) and Variability of Voicing (VoV) obtained from the bispectral distribution, which when used along with the popular prosodic cues helps in drastically reducing the recognition error rate in-volved. An ensemble of three different models has been designed to minimize the false alarms, during word boundary detection, by maximizing the uncorrelatedness in prediction from each model. Finally, the majority-voting rule was used to decide if the given frame encompasses a word boundary. The contribution of the work lies in: (i) Converting word boundary detection into a supervised learning problem (ii) Introduction of two new higher order statistical features (iii) Using ensemble methods to find the best model for prediction. Experimental results for NTIMIT Database shows the efficacy of the proposed method and its robustness to noise added during telephonic transmission. © 2016 IEEE.Item Selfie Detection by Synergy-Constraint Based Convolutional Neural Network(Institute of Electrical and Electronics Engineers Inc., 2017) Annadani, Y.; Naganoor, V.; Jagadish, A.K.; Chemmangat, K.Categorisation of huge amount of data on the multimedia platform is a crucial task. In this work, we propose a novel approach to address the subtle problem of selfie detection for image database segregation on the web, given rapid rise in the number of selfies being clicked. A Convolutional Neural Network (CNN) is modeled to learn a synergy feature in the common subspace of head and shoulder orientation, derived from Local Binary Pattern (LBP) and Histogram of Oriented Gradients (HOG) features respectively. This synergy was captured by projecting the aforementioned features using Canonical Correlation Analysis (CCA). We show that the resulting network's convolutional activations in the neighbourhood of spatial keypoints captured by SIFT are discriminative for selfie-detection. In general, proposed approach aids in capturing intricacies present in the image data and has the potential for usage in other subtle image analysis scenarios apart from just selfie detection. We investigate and analyse the performance of the popular CNN architectures (GoogleNet, Alexnet), used for other image classification tasks, when subjected to the task of detecting the selfies on the multimedia platform. The results of the proposed approach are compared with these popular architectures on a dataset of ninety thousand images comprising of roughly equal number of selfies and non-selfies. Experimental results on this dataset shows the effectiveness of the proposed approach. © 2016 IEEE.Item Harmonic and switching loss analysis of two-level space vector based pulse width modulation schemes(Institute of Electrical and Electronics Engineers Inc., 2017) Krishna M R, P.; Arumalla, R.T.; Figarado, S.; Chemmangat, K.This paper presents the analysis of harmonic and switching loss for two-level space vector based pulse width modulation (PWM) schemes. The harmonic performance of the inverter is evaluated through weighted voltage total harmonic distortion for different modulation indices. To calculate switching loss, analytical expressions for switching instants are generated by comparing the triangular carrier signal with the equivalent modulating signals corresponding to space vector based modulation schemes and switching instants are obtained in terms of Kapteyn series. The losses are compared analytically for three popular space vector based modulation schemes, namely conventional space vector PWM, 60° bus clamping PWM and 30° bus clamping PWM. © 2017 IEEE.Item Feature selection for myoelectric pattern recognition using two channel surface electromyography signals(Institute of Electrical and Electronics Engineers Inc., 2017) Powar, O.S.; Chemmangat, K.Pattern recognition scheme is used for discriminating various classes of hand motion with feature extracted from the surface electromyography signals. However, while using a relatively large feature set for classification process, the computational complexity increases tremendously. To overcome this, the paper implements feature selection technique using wrapper evaluation and four different search methods without significantly affecting the classification accuracy. The performance of the features is tested on surface electromyography data collected from seven subjects, with eight classes of movements. Practical results indicate that using feature selection methods can achieve the same accuracy with lesser number of features. © 2017 IEEE.Item Comparison of pre-processing filters on the performance of sEMG based pattern recognition(Institute of Electrical and Electronics Engineers Inc., 2019) Powar, O.S.; Chemmangat, K.The noise present in the surface electromyography (sEMG) signals is a significant problem in the control of the rehabilitation scheme. Different noise reducing methods have been discussed and considered individually in previous studies. However, there is limited work on the comparison of different noise reduction strategies. To achieve good performance of Myoelectric Control (MEC) system, the selection of filters becomes essential. The vital contribution of this work is a study into the comparison of three denoising methods including Butterworth filter, Weiner filter and Spectral Subtraction (SS) filter that has been used to remove the noise from sEMG signal. Performance evaluation of the three noise reduction methods is done regarding classification accuracy and computation time. The three denoising methods have been validated on the recorded sEMG of seven healthy subjects while performing eight classes of movements from the two muscle positions on the right forearm. The accuracy is compared with four classifiers namely, J48, k-nearest neighbors (KNN), Naive Bayes and Linear Discriminant Analysis (LDA). Results show that the Butterworth filter provides marginally better performance than the other two filters regarding classification accuracy; when computation time is considered SS filter offers significant savings. A visual inspection of the output of the Weiner filter hints at its utility as a muscle activity onset detection tool. © 2019 IEEE.Item A Computationally Efficient sEMG based Silent Speech Interface using Channel Reduction and Decision Tree based Classification(Elsevier B.V., 2020) Abdullah, A.; Chemmangat, K.Silent Speech Interface is one of the promising areas of Human-Computer Interaction research. The surface electromyography based silent speech interface is a technique where the electric activity of facial muscles are used to detect speech. The existing sEMG based SSI techniques use complex machine learning algorithms and too many number of electrodes on the subject's face. It creates inconvenience to the user who might have undergone laryngectomy. More number of electrodes becomes highly invasive to the user, while complex classification algorithms increase the computational cost and prevents real time implementation of sEMG based SSI. Thus the objective of this research work was to develop a less complex and computationally less expensive model to classify words. To achieve this goal channel reduction technique and the use of Decision Tree based classification algorithm was employed. Only the time domain features are used as input to the classification algorithm. The motive was to exploit the advantage of computational ease in extracting the time domain features as compared to the frequency domain features. The sEMG data of the words used in this work are obtained from the complete utterance of the sentences and not by individual utterances of the word. Our algorithm was able to achieve a word accuracy of 95.17% even after applying a channel reduction, thereby allowing us to use only the data of 5 channels, in place of a conventional seven channel setup. © 2020 The Authors. Published by Elsevier B.V.Item A two-stage classification strategy to reduce the effect of wrist orientation in surface myoelectric pattern recognition(Institute of Electrical and Electronics Engineers Inc., 2022) Koppolu, P.K.; Chemmangat, K.The myoelectric Pattern Recognition (PR) collects surface Electromyographic (sEMG) signals using the electrodes placed on the upper limb of the amputee. Then it recognizes patterns in those signals based on the intended limb movement using signal processing and machine learning techniques. The performance of the PR system should be robust against multiple factors, like wrist orientation, muscle force level changes, limb position changes, and electrode shifts. This paper demonstrates how performance is affected by wrist orientation and proposes a method to overcome those effects. A two-stage classification technique with Dynamic Time Warping (DTW) as the classifier, along with features extracted from a three-axis accelerometer and six-channel sEMG sensors, is proposed here. Accelerometer features are used to identify the wrist orientation, and sEMG features are used to classify the various limb movements performed by ten subjects. The performance of the proposed method was measured by classification error and classification accuracy of limb movements. The corresponding results were compared with the state-of-the-art techniques. © 2022 IEEE.Item A Comparison of Different Signal Processing Techniques for Upper Limb Muscle Activity Onset Detection using Surface Electromyography(Institute of Electrical and Electronics Engineers Inc., 2023) Koppolu, P.K.; Chemmangat, K.This work presents the use of real-time experimental Surface Electromyography (sEMG) signals to determine muscle activity of upper limb by detecting the exact onset and offset timings. Various muscle activity detection methods were evaluated, such as Sample Entropy (SEn), Permutation Entropy (PEn), Amplitude Aware Permutation Entropy (AAPEn), and Integrated Profile (IP). The performance of these methods was compared, and it was found that IP detects muscle activity quickly and requires less computation for real-time implementation as compared to other methods. © 2023 IEEE.Item Improved Robustness of EMG Pattern Recognition for Transradial Amputees with EMG Features Against Force Level Variations(Institute of Electrical and Electronics Engineers Inc., 2023) Powar, O.S.; Chemmangat, K.Feature extraction is an essential process for removing the unwanted part and interference of the Electromyography (EMG) signal, and to extract the useful information hidden in it. Inorder to obtain high performance of Myoelectric Control (MEC), the choice of features plays an important role. The studies carried out earlier to overcome force level variation have used features which are redundant, affecting the robustness and the classification performance. This study's main objective is to assess a database's performance consisting of nine upper limb amputee subjects with EMG data recorded at three different force levels when six motions were classified using twenty different time domain features that are frequently found in the literature. Training is carried out at one force level, and the other two unknown force levels are used for testing. Out of the twenty features, the one that is the most stable is displayed for each force level. The results show that root mean square (RMS) feature outperformed other features for training at low and medium force levels, and Wilson amplitude (WAMP) feature for training at a high force level, when compared with the most widely used linear discriminant analysis (LDA) classifier. The average classification accuracy for the nine amputee subjects trained with the RMS feature at low and medium force levels was 42% and 51.78% percent, respectively. For high force level, when trained using WAMP feature, an accuracy of 46.78% has been obtained. The features are verified using histogram plots. This study will help select those features which are not important for robust classification of hand movements. © 2023 IEEE.Item Feature Selection and Ranking in EMG Analysis for Hand Movement Classification(Institute of Electrical and Electronics Engineers Inc., 2023) Chandrika, P.R.; Powar, O.S.; Chemmangat, K.Surface Electromyography has gained tremendous significance in the recent years due to its suitability and reliability in a wide range of applications like automatic prosthetic control, diagnosis of neuromuscular disorders, in robotics and many such fields. Considering such applications, identification of various muscular movements is necessary and hence, EMG pattern recognition is needed. This paper focusses on a generalised EMG pattern recognition of various hand movements. The data from Ninapro Database - 4 has been used for pattern recognition. The database has Surface Electromyogram (sEMG) data of 52 various hand movements. The data was subjected to pre-processing, feature extraction and classification. An average accuracy of 64.87% was obtained for a combination of seven features in the time (temporal) domain, using Linear Discriminant Analysis (LDA) as the classification model. The obtained classification accuracies are compared and discussed with respect to the state-of-the-art literature. © 2023 IEEE.
- «
- 1 (current)
- 2
- 3
- »