Removal of noise from ECG Signals leads to the accurate analysis of potential diseases. ECG Signals are low frequency signals. In this paper, FIR low pass filter have been designed with the help of window techniques at cut off frequency 60 Hz to remove noise from corrupted signal. Additive white Gaussian noise (AWGN) is added artificially to the ECG samples recorded from MIT-BIH database. Comparison of ECG Signal before and after filtering is done on the basis of two parameters i.e. signal to noise ratio and average power. The results are calculated using Gaussian, Bartlett and Hann window based FIR filter.

In this paper, designing of a low pass FIR filter is done by artificial neural network. For this kind of application a different type of model is used in ANN. In this research work, selforganizing map (SOM) algorithm is used to train the Neural Network. The SOM algorithm is based on unsupervised learning technique. We also compare the result of neural network with the result of normal window method. The accuracy of SOM neural network is about 98%.

No abstract

In this paper a low pass finite impulse response (FIR) filter has been designed using artificial neural network. The optimization of the network has been done using generalized regression algorithm. The proposed approach has been compared with rectangular window method. The accuracy of this algorithm is about 98%, that is much higher than multi layer perceptron (MLP) back propagation algorithm.

In this digital world, everything including documents, notes is kept in digital form. The requirement of converting these digital documents into processed information is in demand. This process is called as Handwritten digit recognition (HDR). The digital scan document is processed and classified to identify the hand written words into digital text so that it can be used to keep it in the documents format means in computerized font so that everybody can read it properly. In this paper, it is discussed that classifiers like KNN, SVM, CNN are used for HDR. These classifiers are trained with some predefined dataset and then used to process any digital scan document into computer document format. The scanned document is passed through four different stages for recognition where image is preprocessed, segmented and then recognized by classifier. MNIST dataset is used for training purpose. Complete CNN classifier is discussed in this paper. It is found that CNN is very accurate for HDR but still there is a scope to improve the performance in terms of accuracy, complexity and timing.