Dalhousie University Faculty of Computer Science
CSCI 3162: Digital Media — Assignment 4
Winter Term 2021
due Tuesday, April 6, 23:59 ADT
1. Discrete Fourier Transform: Use the Fast Fourier Transform algorithm to compute the 8-point Dis- crete Fourier Transforms of sequences (1, 0, 1, 0, 1, 0, 1, 0) and (1, 1, 1, 1, 0, 0, 0, 0). Provide detailed calculations with intermediate steps rather than just the results.
2. Audio analysis:
• Use MATLAB to analyze the first few seconds of piano.wav. Try to determine which notes are played. Include the output of spectrogram and carefully explain your reasoning behind the parameter settings that you have chosen. Annotate the plot if you can.
• Use MATLAB to generate the Discrete Fourier Transform of a pure sine wave, and those of the same sine wave quantized using a bit depth of six and a dithered version of the latter. Plot all three magnitude spectra (paying attention to the correct scaling and labelling of the axes) and discuss the differences. From the spectra, can you infer why the dithered version sounds less objectionable than the undithered one? Marks will be based on the interpretability of the plots as well as on the correctness and significance of the conclusions.
3. Deconvolution:
• Generate a blurred, noisy version of the cameraman image as described on the lecture slides.
• WriteMATLABcodethatimplementstheWienerfilterfordeconvolution.DonotuseMATLAB’s
deconvwnr.
• Use the code that you have written the deblur the image, first with the (sharp) cameraman power spectrum, then with that of the building. In each case, compute the sum of squared errors that the Wiener filter strives to minimize.
Submission instructions: Please submit your solutions on Brightspace. Remember that
• All work you submit must be your own.
• AnyquestionsyoumayhaveshouldbebroughtupduringclasstimeorpostedontheBrightspace discussion board.
• You must not share your calculations or code with anyone.
• You must not make use of any code you find on the web.