DIGITAL SIGNAL PROCESSING AND PROCESSOR

This patent describes a new innovative technique for sensing ECG and filtering out the noise in the sensed ECG. This technique makes use of the Spatially Selective Filtering (SSF) and Periodic Component Analysis (PCA). Spatially Selective Filtering is based on the correlation of the input signals. A high level of correlation defines whether a signal component is to be passed by the filter or not. This decision is made by the use of Periodic Component Analysis. The patent describes the algorithm to carry out the denoising of the sensed ECG signal. A sensed ECG signal first undergoes signal conditioning. The conditioned ECG is then broken down into large number of uncorrelated signals. Further these components in the signal are analysed by SSF and PCA for noise components. SSF and PCA remove noise components and the original noiseless ECG signal is reconstructed using inverse of the transform used to decompose the signal initially. Calculation of noise threshold is a ...

The paper concludes the work done so far the denoising of ECG signal and authentication. The cardiac muscles is generating the cardiac signals that representing by electrocardiogram. An ECG signal is noise free and it is used to analysis and identifying the person. It is necessary to denoised the loud signal using various techniques like wavelet transform, Kalman Filter, FIR filter, etc. noisy signals are decomposed and noisy components appear in coefficients of detail. The threshold applies to corrupt bands. Using inverse technique the detailed sub-bands are reconstructed. The feature is extracted from denoising of ECG signals and classified the persons. Various classifiers are used and compared to existing methods. K-NN, ANN, SVM are the common techniques of classification used. The future work mainly concentrates on developing an algorithm for de-noising and accurate authentication. Plagiarism check was done on Plagiarism Checker. The report for the same is as follows:

One of my classmates, sahil- who had already worked on dsp processor helped me understand the dsp processor and how much better it is from the normal processor. It was  particularly interesting to know the difference. There are several real time applications of dspp processor.  Code Composer Studio (embedded C language) is use to program the processor. We performed basic arithmetic operations such as addition and multiplication by using this processor. The operations such as circular shifting of data can be done by using this in one clock cycle.

This experiment taught me that: Stability of the filter is decided by the placement of analog poles on the s-plane. If the poles lie on the left-side of the plane, filter is completely stable. There are two methods to design IIR filters: Bilinear Transformation Method(BLT) and Impulse Invariant Method(IIM). IIM method gives many-to-one mapping.Hence it is not suitable for HPF and BPF filter design. Butterworth filter design was done using Scilab and MATLAB software.

Overlap Add and Overlap Save method are used for filtering long data sequences. Such a long data sequence would need a very high order FFT- which is not feasible and the process can only start after we've obtained the entire data sequence which leads to a delay. The long data sequence is broken into smaller parts and performed FFT analysis. It is then fitted with one another, the overlapped portion is discarded in Overlap Save Method whereas it is added in OAM to get required output sequence.

Fast Fourier Transform(FFT) A   Fast Fourier transform  ( FFT ) is an algorithm that samples a signal over a period of time (or space) and divides it into its frequency components. FFT is manually verified and performed using flowgraphs. FFT is much faster than DFT and more efficient as it requires less number of real and complex additions and multiplication. The computation is considerably reduced from M2N2 to MNlogMlogN

Discrete Convolution Convolution  is a common mathematical way of combining two signals, input, and impulse- to form a third signal, the output signal. The condition for the length of the output of Linear Convolution is greater than or equal to one less the sum of the length of an input signal and impulse response signal(N>=L+M-1). Circular convolution gives an aliased output. Correlation Correlation gives a measure of similarity between two data sequences. In this process, two signals are compared and the degree to which the two signals are similar is computed. In cross-correlation, the output sequence gets affected when the input signal is delayed. Autocorrelation The correlation of a function with itself is called its autocorrelation. In Autocorrelation, the delay in the input signal would make no difference in the output signal i.e Autocorrelation of x[n-1] is same as autocorrelation of x[n].