KUMAR M
@chettinadtech.ac.in
Associate Professor/Electronics and Communication Engineering
Chettinad College of Engineering and technology
RESEARCH, TEACHING, or OTHER INTERESTS
Engineering, Signal Processing, Automotive Engineering, Information Systems
Scopus Publications
Scopus Publications
Janani S, Akhil Hassan G, Madhankumar S, and M. Arun Kumar
IEEE
Speech is the primary means of human communication. Speech has the potential to be a more effective interface than keyboards and pointing devices. A speech interface would support a wide range of useful applications, including phone directory assistance, "hands busy" medical applications, office dictation devices, etc. This has stimulated research on Automatic Speech Recognition (ASR). Speech recognition has grown quickly in recent years thanks to advancements in parameter extraction tools for spoken signals. Automatic speech recognition (ASR) in noisy environments is still a challenge since there are many possible environmental distortions and it is hard to accurately compensate for them. Low recognition performance in noisy environments is mostly caused by mismatches between training and test circumstances. Noise and distortion are the main issues limiting communication systems. Thus, the modeling and removal of distortion and noise effects have been the cornerstones of communications theory and practice as well as signal processing. Noise reduction and distortion removal are important difficulties in voice recognition, picture processing, medical signal processing, radar, sonar, and any other application where the signals cannot be isolated from noise and distortion. There is noise in almost all auditory environments. In applications related to speech, sound recording, telecommunications, speech recognition, and human machine interfaces, the signal of interest typically speech is usually contaminated by noise coming from multiple sources. Today’s speech recognition systems' acoustic modeling components are primarily based on the Hidden Markov Model (HMM).This HMM is well renowned for being a successful model for speech signals and is the most commonly used paradigm for speech recognition. The primary reasons for the reduction in performance are typically the non-native speaker’s lack of fluency and the phonetic discrepancies between the target language and mother tongue. Requiring front-end signal processing is necessary to make voice recognition systems feasible. This is because better signal feature extraction leads to better recognition performance. Voice Activity Detection (VAD) and Speech Enhancement Algorithm (SEA) are used in a preprocessor to improve Recognition Accuracy (RA) in ASR. In order to increase the percentage of ASR’s Recognition Accuracy (RA), this thesis looks at a novel approach for speech enhancement and voice activity detection techniques. A hybrid approach was also proposed to increase the Recognition Accuracy % in different noisy scenarios. In comparison to the proposed VAD and Speech Enhancement algorithms, the recommended hybrid algorithm performs better for variable noise levels at varying Signal to Noise Ratio (SNR) values. Better RA was seen for the proposed hybrid technique in the presence of station noise (86.28%) at different noise levels.
M. Arun Kumar and Arvind Chakrapani
IEEE
One of the most vital parts of the human body is the heart, which circulates blood throughout the body and transports oxygen, nutrition, and waste products. However, the shift in lifestyle and environmental factors results in an aberrant heart's ability to beat. Cardiovascular diseases (CVDs) are the leading cause of death worldwide and the most prominent health concern today, impacting people of all ages. Heart and blood vascular illnesses are grouped as CVDs. The two primary subtypes of cardiac arrhythmias (CAs), a category of cardiovascular diseases (CVDs), are atrial and ventricular. According to WHO estimates, around 61% of individuals globally have CVD. In percentage terms, the disease affects 15%, 10%, 5%, and 5% of the population. The benefits of a wavelet-based VS method are merged with WF in the hybrid VS/WF technique. EMG interference and power-line interference are two examples of noise sources used to gauge the efficacy of the hybrid VS/WF technique. Numerous quality indicators are also looked at. VS/WF hybrid's performance is compared to well-known thresholding techniques including Visu Shrink, Global SURE Shrink, and hybrid threshold approach. The latter of the three threshold techniques Hybrid, Global SURE Shrink, and Visu Shrink is the best.In order to evaluate how well TNN works when supervised learning techniques are applied, three optimization approaches Gauss-Newton, Newton Raphson, and Leven berg Marquard are used. The de-noised ECG data undergo additional processing in order to extract characteristics. A number of domains, including Time, Frequency, and Time-Frequency (Wavelet) domains, are used to extract the characteristics. Auto-regressive (AR) coefficients are extracted in the time domain. While relative wavelet energy is extracted in the wavelet domain at various decomposition levels, Power Spectral Density (PSD) values are recovered in the frequency domain. These characteristics are used to construct an Artificial Neural Network (ANN) that is fully connected and has an accuracy performance rating of (96.85%) for classifying arrhythmias.A more effective de-noising, feature extraction, and classification model based on Conventional Neural Networks (CNN) are also developed. Compared to ANN, the performance is judged as having (99.2%) accuracy. Therefore, the suggested CNN model is helpful to physicians in reaching the ultimate diagnosis of atrial fibrillation (AFIB), atrial flutter (AFL), and ventricular fibrillation. It incorporates de-noising, feature extraction, and classification VT with Ventricular Fibrillation (VFL).
T. Gobinath, S. Sathish Kumar, P. Ramya, M. Kumar, L. Anbarasu, and N Padmavathi
IEEE
This paper explores the synergistic integration of robotics and Internet of Things (IoT) monitoring within industrial settings, employing Deep Residual Networks (Deep ResNets). The convergence of robotics and IoT has the potential to revolutionize various industries by enabling efficient data collection, real-time analysis, and informed decision-making. Deep ResNets, known for their exceptional feature extraction capabilities, further enhance the fusion of these technologies. In this study, we delve into the collaborative framework of robotics and IoT, elucidating how Deep ResNets bolster the process. We discuss the benefits of this integration, including optimized resource allocation, predictive maintenance, and adaptable production systems. Through empirical analysis, we demonstrate the efficacy of Deep ResNets in handling complex data from integrated robotics and IoT sources. Ultimately, this research sheds light on the transformative impact of amalgamating robotics, IoT monitoring, and Deep ResNets, paving the way for intelligent and responsive industries.
Mebin.K. B, T. Loganath, S. Ariharan, and M. Arun Kumar
IEEE
Many patients have been diagnosed with serious cardiac diseases in the recent years. Even teenagers suffer from heart disease and this might affect anyone due to unhealthy diets, lack of physical activity, having more junk foods, etc. On the other hand, heart disease can be caused due to smoking and consuming alcohol. Hence heart disease must be monitored as earliest as possible and needs to be treated before the condition gets severe, which forms the main motive of this research. Real-time ECG monitoring is performed in this work using flexible capacitive electrodes through a wearable smart T-shirt. In this study, flexible electrode is created which can be placed in the T-shirts and attached onto the patient's body who is vulnerable to heart disease. The heart disease is diagnosed at an earlier stage using this suggested system and the patient can be treated immediately based on the acquired ECG signal. The ECG signal is captured using AD8232 ECG module and the captured signal is monitored in the serial plotter for obtaining the ECG waveform.
Arun Kumar M. and Arvind Chakrapani
Public Library of Science (PLoS)
Electrocardiograms (ECG) are extensively used for the diagnosis of cardiac arrhythmias. This paper investigates the use of machine learning classification algorithms for ECG analysis and arrhythmia detection. This is a crucial component of a conventional electronic health system, and it frequently necessitates ECG signal reduction for long-term data storage and remote transmission. Signal processing methods must be used to extract the function of the morphological properties of the ECG signal changing with time, which is difficult to discern in the typical visual depiction of the ECG signal. In biomedical research, signal processing and data analysis are commonly employed methodologies. This work proposes the use of an ECG arrhythmia classification method based on Fast Fourier Transform (FFT) for feature extraction and an improved AlexNet classifier to distinguish the difference between four types of arrhythmia conditions that were collected from records. The Convolutional Neural Network (CNN) algorithm’s results are compared to those of other algorithms, and the simulation results prove that the proposed technique is more effective for various parameters. The final results of the proposed system show that its ability to find deviations is 20% better than that of traditional systems.
M. Arun Kumar and G. Aravindh
IEEE
The radiological characteristics of water will generally refer to the chemical, physical, biological, and radiological characteristics of water. The water quality should be strictly maintained to ensure the survival and growth of aquatic lives. Henceforth, in order to maintain the water quality, autonomous system should be implemented for monitoring Aquaculture by implemented IOT Technology is proposed in this paper. The proposed system will contain different sensors like Temperature Sensor, Water Level Sensor and the Sensor PH Sensor. Temperature, PH and Water level will play a major role in ensuring the water quality. The changes in these parameters may adversely affect the quality of the water and also the aquatic lives. GSM is used to communicate to the respective mobile phone of the user. Temperature Sensor is mainly used to detect the temperature changes and if the temperature rises above the normal value, the DC Motor will move to ON condition. The PH Sensor is used to monitor the PH value of the Soil continuously. The level of the water is monitored by using the Water level sensor and if level of the water exceeds the normal value, then the motor will be switched ON. The sensed data will be then transferred to the respective mobile phone through GSM. The GSM will send message to the Mobile when the sensors detect any abnormal value. LCD is used to display all the sensor values.
S. Jayachitra, M. Arun Kumar, G. Aravindh, R. Sangeetha, and P. Sasikala
IEEE
Seismic wave is the wave of energy that travels via the earth as a consequence of an earthquake or a volcano that occurs with low-frequency acoustic energy. By analyzing the Seismic earthquake patterns it is possible to find whether the earthquake will occur or not. In this project, the seismic data are analyzed by a Fuzzy Logic Controller. But the seismic data are affected by some unwanted artifacts. The de-noising method has the capacity to reduce the artifacts. Here the Contourlet transforms scheme gives better noise suppression in seismic images and the signal to noise ratio is improved. Further, the de-noised seismic image is segmented using Wavelet transforms for feature extraction. The extracted features can be analyzed by fuzzy Logic Controller which shows whether the seismic image is normal or abnormal. This analysis gives better results in the earthquake analysis and also in the construction of sensitive power plants.
G. Aravindh and M. Arun Kumar
IEEE
To avoid the traffic congestion and to make the efficient parking slot, it is necessary to find the parking slot within close proximity due to increased number of vehicles. In the existing system, the microcontroller was used. The main problem in the microcontroller is to register the mobile number in the central server which is located in the parking lot. The Raspberry Pi is used to overcome the problem by developing a new application in the android mobile phone. So any number of users can find the parking lot without any registration. It has vechicle detection units in the guidance of parking lot. The processor unit helps to transmits the availability of parking space from vehicle detection units based on the dectection results and android mobile for showing the available parking space in a each floor of the parking lot..
G. Aravindh and M. Arun Kumar
IEEE
The children diagnosed with hearing disabilities will lack the ability to recognize a word or sentence when it is orated. The Computer aided language learning system will remain as a boon for the children ith hearing impairment. The solution for this challenge is articulatory speech learning that has been performed by using Electropalatography(EPG) methodology, which is usually a task based language learning. The dataset combines speech and EPG of English vowels and consonants. Different statistical classifiers are investigated to recognize the vowel of electropalatography signals. The support vector machine, quadratic discriminate analysis algorithms and Linear discriminate analysis are used to classify the vowels and consonents. SVM and QDA Algorithms have been implemented to analyze the articulatory synthesis and speech synthesis, which will together decide the intonation contour of the speech utterances and tongue gestures.
M. Arun Kumar, S. Jayachithra, G. Aravindh, and M. Bhuvaneswari
IEEE
Robots will usually interact with the people directly, and hence it is very important to find an easier way for user interface. Only few robotic systems are user interfaces that possess the ability of controlling the robot by natural means, while issues such as manipulation and navigation in the environment have been focused primarily by earlier works.To promote a beneficial solution to this requirement, a system has been implemented through which the user can give commands to a wireless robot using gestures. With the help of this method, the robot can be navigated by the user by gestures using fingers, and thereby providing a way for interaction with the robotic system. By using image processing, command signals are generated from those gestures. Those command signals are then passed to the robot to navigate it in the specified direction.
M. Kumar and V. E. Jayanthi
Springer Science and Business Media LLC
M. Kumar and V. E. Jayanthi
Springer Science and Business Media LLC
Kshitij Dwivedi, Nitin Singh, Sabari R. Shanmugham, and Manoj Kumar
Springer Singapore
Predicting where humans look in a given scene is a well-known problem with multiple applications in consumer cameras, human–computer interaction, robotics, and gaming. With large-scale image datasets available for human fixation, it is now possible to train deep neural networks for generating a fixation map. Human fixations are a function of both local visual features and global context. We incorporate this in a deep neural network by using global and local features of an image to predict human fixations. We sample multi-scale features of the deep residual network and introduce a new method for incorporating these multi-scale features for the end-to-end training of our network. Our model DeepAttent obtains competitive results on SALICON and iSUN datasets and outperforms state-of-the-art methods on various metrics.
Nitin Singh, Manoj Kumar, P.J. Mahesh, and Rituparna Sarkar
IEEE
The rapid development of camera in hand-held devices and the emergence of social media has led to an uprise in capturing self-portrait images. Augmenting these images for beautification or applying special effects to mimic DSLR camera has become a popular practice. Most of these effects require separation of foreground from background where the effect can be applied solely on background. To employ such effects on portrait (upper half of human body) images, a pixel-accurate segmentation is imperative. In this paper, we propose an effective method of fast depth aware CNN based portrait segmentation from monocular images. The proposed method is capable of being deployed on mobile phones, within the constraints of time and memory. On the segmented images, we demonstrate the application of bokeh effect, which blurs out-of-focus regions. We experiment with different combinations of state of the art encoder and decoder networks for segmentation and infer that our proposed method can improve the inference speed by 76 ms on mobile device while maintaining an accuracy of 97.0 %.
Sushil Kumar, Millie Pant, Manoj Kumar, and Aditya Dutt
Springer Science and Business Media LLC
Due to the complexity of underlying data in a color image, retrieval of specific object features and relevant information becomes a complex task. Colour images have different color components and a variety of colour intensity which makes segmentation very challenging. In this paper we suggest a fitness function based on pixel-by-pixel values and optimize these values through evolutionary algorithms like differential evolution (DE), particle swarm optimization (PSO) and genetic algorithms (GA). The corresponding variants are termed GA-SA, PSO-SA and DE-SA; where SA stands for Segmentation Algorithm. Experimental results show that DE performed better in comparison of PSO and GA on the basis of computational time and quality of segmented image.
M. Arun Kumar and G. Arun Francis
IEEE
Survey on non-uniform cache design concludes the paper that may be a future approach to an outsized range of core processors. Cache may be a memory in between the processor and also the main memory. A smaller temporary memory that manages the main memory locations and access time thereby will increase speed throughout execution time With the trend of transient error rate, its turning into necessary to forestall transient errors and supply a correction mechanism for hardware circuits, specially for SRAM cache recollections. Caches measure the most important structures in current microprocessors and, hence, square it measures most prone to the transient errors. This paper at first exploits the same tag bits to enhance error protection capability of the tag bits within the caches. Once information measure fetched from the most memory, it's checked if adjacent cache lines have identical tag bits as those of the info fetched and this paper incorporates a thorough discussion concerning cache and varied mapping technique. Then we have a tendency to shift our focus on cache optimizations and discuss the motivation for doing this on the same, followed by the completely different improvement techniques. any to avoid varied classes of cache misses we have a tendency to discuss completely different sorts of cache technique to reach higher performance. Lastly we have a tendency to discuss few open and difficult problems featured in varied cache improvement techniques.