DSPexpt

The aim of this experiment was to perform various operations on DSP Processor. We used TMS320F2855 processor to perform those tasks. We used Code Composer Studio to program to processor. We performed Arithmetic operations like Addition, Subtraction, Multiplication and logical Operations like AND, OR, NOT.The change in the register values was observed. 

We learned that FIR filter design is done by window method and frequency sampling method. The Windowing method is used for truncation of the signal with respect to window functions like rectangular, hamming , hanning, bartlett, Blackmann. Frequency sampling method samples the DTFT signal at frequency of (2*pi*k/N). We wrote a script on scilab to design these filters and obtained the graphs.

IEEE Paper : DSP Implementation of Voice Recognition Using Time Warping Algorithm Department of Electronics Engineering, NED University of Engineering & Technology, Karachi Date of Conference - 27 Aug 2005 Date added to IEEE Xplore - 12 Nov 2007. Review: The paper tells about voice recognition algorithm using Time Warping. Voice recognition is a process of an automatic system to identify the speech. Paper tells about voice recognition using Cepstral analysis & DTW of set of five words. The developed models provide tools to record, filter and analyze different voice signals and compare them with output sample. Voice signal analysis is done using various operations like pre-emphasis, normalizing, windowing, cepstrum analysis and recognition of the input word. Recognition block uses dynamic time warping algorithm. The software used for implementation are MATLAB, Simulink, Softscope (Data Acquisition Toolbox ). Hardware used is C67x consisting of three main parts: CPU, ...

The aim of this experiment was to design Analog and Digital Butterworth LPF and HPF using IIM and BLT. The scilab program was written to accept inputs as As, Ap, Fp, Fs, F(sampling frequency) and the magnitude spectrum was plotted to get the output. In BLT observed parameters are nearly matching to get exact values order must be increased. Digital LPF poles lie inside the unit circle this shows that both analog and digital butterworth filters are stable. In IIM method as frequency mapping is many to one, this method is suitable to design LPF but not HPF or BPF.  The experiment took long time to get completed. Because we didn't knew about HPF designing from IIM method can be performed or not and we were repeatedly trying to get output spectrum of HPF.

This was 4th experiment. In this experiment we perform filtering of Long Data Sequences using OAM and OSM. To perform a c program was written and executed and output was found as desired. Overlap Add Method        In this method input signal x(n) is decomposed into no. of L point sequences using formula N=L+M-1, where M is length of h(n) and N is assumed to be Radix2 value. Then we find linear convolution between every decomposed signal and h(n) using FFT algorithm. The Inverse FFT is then performed on all output signals and the outputs are added to get final result. Overlap Save Method       In this method input signal x(n) is again decomposed into L point sequences using formula N>=L+M-1, where M is length of h(n) and N is assumed to be Radix2 value. Here every decomposed signal is begins with (N-L) last values of previous signal. Then we find linear convolution between every decomposed signal and h(n) using FFT algorithm. The Inverse FFT is...

Fast Fourier Transform- In this experiment, we learned to write the c program to perform FFT of  4 point and 8 point signals. We had learned about FFT algorithm in class before coming for practical. So we had basic idea about FFT flowgraphs in theory DFT operation is faster but in practical FFT algorithm is faster. Generally Radix-2 length signals are preferred in FFT algorithm, but we can reduce any no. of length of signal into prime no. base FFTs. For longer sequences, FFT takes more time for output computation as the no. of stages in the Flowgraph increase.