Audio Processing Exercise

Transcription

1 Name: Date : Audio Processing Exercise In this exercise you will learn to load, playback, modify, and plot audio files. Commands for loading and characterizing an audio file To load an audio file (.wav) in MATLAB use the following command >> [x, fs, nbits] = wavread( 'path\file_name'); Here x is the array created containing the sampled data that s been loaded. To find the total number of samples, call it (N), take the length of x >>N = length(x) fs is the sampling frequency. (Sampling frequency is the number of samples per second) nbits is the number of bits used to represent each sample. The total number of bits used by the sound file, call it (B) is found from B = Number of samples Bits/sample or >>B = N nbits The time interval between samples is known as the sampling period Ts, and is related to the sampling frequency by; >>Ts = 1/fs The total time taken to record, or play, all the samples is found from >>T = N Ts

2 Recall: The steps required to digitize an electrical signal representing a sound wave Original Analog Signal Sampled Discrete-Time Signal Discrete-Time Quantized Signal Discrete-Time Quantized Coded Signal

3 Using MATLAB Load the audio file tutor1.wav, download from the course web site, and find some parameters. %Loads the sequence >> N = length(x) %Total number of samples N = >> fs % Sampling frequency fs = >> nbits % number of bits per sample nbits = What is the binary range of numbers, starting at zero, can be used to represent amplitude levels based on your computed value for nbits? >> B = N*nbits %Total number of bits used by the file B = What is the file size in Kbytes? >> Ts = 1/fs % Time between samples Ts = >> T = N*Ts %Total time taken T =

4 Play the loaded sequence with MATLAB The command sound can be used to play a sequence as follows. sound(x, f) Where x is the sampled data and f is the frequency at which we need to play the audio Example: Take the audio file tutor1.wav load it with MATLAB and play it with MATLAB as follows >> sound(x, fs) % Plays the audio with sampling frequency To play with different frequencies first find out the current playback frequency >> fs fs = If you increase the sampling frequency the total time T taken will be decreased, Ts = 1/fs and T = N*Ts therefore the file will play back faster. If you decrease the sampling frequency the total time taken will be increased and the file will play back slower. Play the sequence at the listed values for fs and see if you can notice the difference between frequencies. >> sound(x, 8000) >> sound(x, 9000) >> sound(x,10000) >> sound(x,11025) >> sound(x,12000) >> sound(x,13000) >> sound(x,14000)

5 Plotting the sound wave Example: Plotting sample index vs. amplitude >> plot (x) Example: Plotting time vs. amplitude To plot against time you need to convert the sample positions into units of time. Computing the length of x will yield the total number of sample positions >> N = length (x); Now to convert to time you want to create an array from 0 to the maximum time, since you will be including 0, the time will run from 0 to (N-1) positions. Since fs is the number of samples taken per unit time, the time between each sample is 1/fs. Then the time t is: >> t = 0 : 1/fs : (N-1)/fs; >> plot (t, x)

6 One characteristic of a sound wave you will be interested in is its frequency content. In other words you are interested in the frequencies that make up the signal. A time domain signal can be converted to a frequency domain signal using a Fast Fourier Transform (FFT), which returns the Discrete Fourier Transform (DFT) of the signal. If the DFT is plotted you will see the spectrum of the signal. This exercise will not cover the specifics of the FFT, however MATLAB has a built in function to apply an FFT to a signal. y = fft(x,n); n is the number of points in the FFT, if left out the length of y = length of x. This command returns the DFT of x. Try it: >> y = fft(x); After taking the FFT, the new array y will contain mirror, real, and imaginary information, this is the nature of the FFT. When plotting the spectrum we will be concerned with only the magnitude of the signal which contains the frequency and amplitude information, so we must use the abs() command. >> y = abs(y); To convert the horizontal scale from sample count to frequency, create a new array f >> f = [-fs/2 : fs/(length(y)-1) : fs/2]; To center the plot, mirror images of frequency, at zero use the fftshift command >> y = fftshift(y); The two previous commands are based on knowledge of how the FFT is preformed and the resulting information. You will learn about the details in your signal and systems course. Now plot the spectrum >> plot(f,y)

7 Exercise: 1. Using the windows sound recorder, record yourself counting from 1 to 10, then use MATLAB to play it back and compute; N, fs, nbits, B, Ts, and T. 2. On one page (i.e. using subplot), plot a. The wave file as a function of samples. b. The wave file as a function of time. c. The wave file as a function of frequency. 3. Turn in this exercise with answered questions, and your computed values for; N, fs, nbits, B, Ts, and T and plots a-c.