Fast Fourier Transform in Python

Two 16-bit sampled (@44100Hz) mono audio records (music) were provided. You had written a Python code for implementation of Radix-2 Decimation in Frequency (DIF) type Fast Fourier Transform. You have also applied 256-point Radix 2 DIF-FFT code on one of the audio records at three spots of the song (e.g. at 10th, 20th and 30th second of the record). 1. Draw three graphs that shows variation of the Fourier convolution between the 256-point length audio record samples at three spots and the overall audio record by shifting the 256-point length window. 2. Draw additional graphs by zooming in ±200 convolutional values range around the window location where 256-point sample is convolved with itself (for each of three graphs). Mark the maximum convolution value point. 3. Can you propose a feasible improvement on this method (FFT convolution) for recognition or similarity comparison of longer audio samples in 100milliseconds range? (explanation only, block diagram or algorithm can be added if necessary. Any method that is used for the same purpose in the scientific literature can also be given) Please includes comment lines in the code. Plots (graphs) and explanations must be included in the report file.