DSPP

The final experiment of the semester was a group experiment- Our task was to pick a domain under signal processing. Our group(Devashish More , Divya Khandkar, Tanvi Joshi, Aditi Kajrekar and Aarohi Karnik) chose speech recognition. Each of us have reviewed a paper and a patent: Patent -  SYSTEM AND METHOD FOR DETERMINING RECIPIENT OF SPOKEN  COMMAND IN A CONTROL SYSTEM Authors -  JoBargetzi Fre ,    Seferian Ara,    Stene Josh,  Labosco Mark  Link to patent review Link to patent                Paper - FRAME LEVEL ENTROPY BASED OVERLAPPED SPEECH DETECTION AS A PRE- PROCESSING STAGE FOR SPEAKER DIARIZATION    Authors - Oshry Ben-Harush, Hugo Guterman ,Itshak Lapidot                 Link to paper review                   Link to paper We also worked on our own appli...

After designing an FIR Filter using Window function our next task was to design an FIR filter using Frequency Sampling Method(FSM). The filter was designed by coding in C language in Scilab. The code was executed and the values of Pass and Stop band attenuation were verified using the graph which was obtained.

After performing IIR Filter Design it was time to start FIR filter Design. In this experiment we have designed an FIR Filter using various Window functions. Depending on the values of stop band attenuation the appropriate window function chosen like Bartlett, Hamming, Hanning and Rectangular. After execution of code the values of pass band and stop attenuation were verified from the graph.

After designing Butterworth filtwr there was another IIR filter that was to be designed and that was Chebyshev Filter. The filter was designed by coding in SCILAB.  Both Low Pass and High Pass Chebyshev filters were designed and the results were verified by manual calculations. Increasing  the sampling frequency increased the  accuracy of the results obtained.

After learning DFT and FFT techniques it was finally time  to start designing filters. We started with IIr filter  design. The first filter that we were going to design was a Butterworth Filter. Butterworth filter  was designed by using SCILAB. The results were verified by running the program and manually solving  using the same set of values. Results for both Low Pass and High Pass Butterworth filter were verified.

Overlap Add and Overlap Save Methods are used to process real time signals. Both OAM and OSM breakdown the signal into segments , process each segment and then combine the segments into one final output signal. OAM and OSM methods use FFT because high computational speed is required. In the experiment OAM and OSM were applied on a 13 point sequence. Because of their speed OAM and OSM are preferred for real time processing.