Vishvjit Thakar

Physical activities recognition from ambulatory ECG signals using neuro-fuzzy classifiers and support vector machines

Abstract The use of wearable recorders for long-term monitoring of physiological parameters has increased in the last few years. The ambulatory electrocardiogram (A-ECG) signals of five healthy subjects with four body movements or physical activities (PA)—left arm up down, right arm up down, waist twisting and walking—have been recorded using a wearable ECG recorder. The classification of these four PAs has been performed using neuro-fuzzy classifier (NFC) and support vector machines (SVM). The PA classification is based on the distinct, time-frequency features of the extracted motion artifacts contained in recorded A-ECG signals. The motion artifacts in A-ECG signals have been separated first by the discrete wavelet transform (DWT) and the time–frequency features of these motion artifacts have then been extracted using the Gabor transform. The Gabor energy feature vectors have been fed to the NFC and SVM classifiers. Both the classifiers have achieved a PA classification accuracy of over 95% for all subjects.

Discrete time output feedback sliding mode control algorithm for chattering reduction and elimination

This paper presents discrete time sliding mode control which uses multirate output feedback and power rate reaching law. The method proposed here uses only output samples and past control input for switching function evaluation and sliding mode control. The discrete time version of power rate reaching law has been obtained and then output samples based algorithm has been developed. It has been shown that it is possible to reduce chattering by suitably choosing the reaching law parameters. It has also been demonstrated that it is possible to avoid chattering completely for special case of controller parameters using only output information.The simulation study carried out on numerical example reveals effectiveness of the proposed algorithm.

Discrete time sliding mode control with application to magnetic levitation system

This paper presents an interesting application of magnetic levitation system using discrete sliding mode control. There is a limited literature available for sliding mode control as applied to magnetic levitation system. In this work a model of linearized magnetic levitation system having small magnetic disc and cylinder with light weight is considered. As a part of review PD control and full state feedback with pole placement and Kalman estimator are presented. A discrete sliding mode control with multi-rate output feedback is then investigated for the present application. To mitigate chattering, discrete time power rate reaching based algorithm is applied. Numerical result for nonlinear system is also shown. In all the cases, inherently unstable system is shown to exhibit stability and stable initial condition response compared to the uncompensated system.

Implementation of derivative based QRS complex detection methods

In this paper QRS complex detection algorithms based on the first and second derivatives have been studied and implemented. The threshold values for detecting R-peak candidate points mentioned in previous work have been modified for accuracy point of view. The derivative based QRS detection algorithms have been found not only computationally simple but exceptionally effective also on variety of ECG database that includes highly noisy and arrhythmic ECG signals. This is indicated by an average detection rate of over 98% obtained through the modified threshold values even for the challenging ECG test sets.

Speed improvement in image registration using maximum likelihood based mutual information

One of the important steps in image fusion is image registration. The process of determining the spatial transformation that maps the points in the target image to the points in the source image is known as image registration. Various image registration approaches can be classified as area, feature and transform domain based. Choice of approach depends on image contents and application. In case of area based approach, during the image registration process similarity measure (also known as similarity metric) is required to measure the similarity between two images. Various similarity measures have been reported such as sum of squared difference, sum of absolute difference, cross correlation, normalized cross correlation and Mutual Information (MI). For multimodal image registration MI is more appropriate. The computation time for MI is challenging for large images for example, satellite images. Various techniques are found in literature to compute MI. One of the techniques is maximum likelihood mutual information (MLMI), which estimates MI between two random variables. In this paper, we have performed image registration using MI as a similarity measure. In this case MI is computed using two approaches; one is MLMI and second is histogram based. Computation time for image registration process on various images is observed and compared for both approaches. It is shown that computation time of IR based on first approach is less than the second approach.

Application of radon transform for fast image registration

One of the important steps in image fusion is image registration. The process of determining the spatial transformation that maps the points in the target image to the points in the source image is known as image registration. Various image registration approaches can be classified as area, feature and transform domain based. Choice of approach depends on image contents and application. Area based approach requires more computation time, specially for large images such as satellite images, while feature based approach may not be accurate if significant features are not available in the images. In this paper authors have used the rotation and translation invariant properties of radon transform to find the amount of rotation and translation required to perform registration, i.e. to align the images. Simulation results are shown for different images, with different amount of rotation and translations, to show the accuracy and reliability of the method. Again noise level is also varied, to observe the robustness of the method to noise. The required average computation time is in seconds, depending on the size of images.

Scale Invariant Feature Transform Based Image Matching and Registration

This paper presents Image matching and registration method that is invariant to scale, rotation, translation and illumination changes. The method is named as Scale Invariant Feature Transform (SIFT). This algorithm will detect and describe image features such as contours, points, corners etc. SIFT descriptors are the characteristic signature of the feature. The features calculated from the image to be registered should be distinctive and then it can be matched. It can be useful in object recognition, image mosaicing, 3 D reconstruction and video tracking. The simulation results shows that this algorithm works well in all types of cases having scale and rotation difference, it also register the object having occlusion and clutter background.

Combining accelerometer data with Gabor energy feature vectors for body movements classification in ambulatory ECG signals

Wearable ambulatory ECG (A-ECG) signals obtained using wearable ECG recorders inherently contain the motion artifacts due to various body movements of the subject. Classification of four such body movement activities (BMA) - left arm up-down, right arm up-down, waist twisting and walking-of five healthy subjects has been performed using artificial neural networks (ANN). The accelerometer data and the Gabor energy feature vectors have been combined to train the ANN. The overall BMA classification accuracy achieved by the ANN classifier is over 95%.

Word boundary detection for Gujarati speech recognition using in-ear microphone

This paper addresses end point detection algorithm, which is most important and critical part of speech recognition system. Different approaches for endpoint detections are explained and compared to Gujarati words. Speech end point detection is required to find speech segment edge points in the presence of the background noise. It isolates words from the silent portions and improves overall accuracy of the speech recognition system. Because the silence portion is mostly affected by the noises [1]. In this work speech data is collected with the in-ear microphone. Results suggested that it gives a high recognition rate compared to conventional recording methods, i.e. by keeping the mic outside mouth. Word boundary detection is also useful to separate out the different words form sentences. The end point task is quite difficult in real time, because of low signal to noise ratio in most of the situations. Detected words are then passed to the feature extraction block.

STUDY OF SPECTRAL CHARACTERISTICS OF MOTION ARTIFACTS IN AMBULATORY ECG BY ANALYZING PCA RESIDUAL ERRORS

In this paper, the spectral characteristics of motion artifacts occurring in an ambulatory ECG signal have been studied using principal component analysis (PCA). The PCA residual errors characterize the spectral behavior of the motion artifacts occurring in ambulatory ECG signals. The ECG signals have been acquired from Biopac MP-36 system and a self-developed wearable ECG recorder. The performance is evaluated by power spectral density (PSD) plots of PCA residual errors as well as statistical parameters like mean, median and variance of PCA errors. The PSD plots clearly indicate that the peak frequency of the motion artifacts occurring due to various body movements (like left and right arms up–down, left and right legs up–down, waist twist, walking and sitting up–down) is located around 20–25 Hz against the ECG peak frequency around 5–10 Hz.