|
|
Laboratory Project M4: Creating Sound Effects |
|
Abstract-You will create script and function files to implement various sound effects based on modifying samples or a sound's spectrum. You will also create a unique sound effect of your own design.
I. Preparation
Read Matlab Primer Chapter 5 pages 5-1 to 5-13, and 5-16 to 5-17.
II. Learning Objectives
1) Use a switch command
2) Use for loops
3) Write function files
4) Use fft (Fast Fourier Transform) command to create spectrum of sound
5) Understand the effects of modifying a spectrum
III. Procedure
A. Lab Report
In this lab, you will write several script and function files. At the conclusion of the lab, send your TA an email whose body is the contents of all these script files and other information as specified below. That is, put all the script and function file contents together, and put them in the body of the email, unless the TA instructs you to do otherwise. Use comments in your files to identify them and how they work.
Show your files (lab report) to your TA before leaving the lab.
B. Sound Effects based on Magnitudes of Samples
Create a script file called clipping.m for this part of the lab. This script file will process a sound waveform based on the values of the samples in the waveform and apply one of several different sound effects, depending on the value of a string variable called effect. Write clipping.m to do the following:
1) Load the handel sound, (which by default will be placed in the variable y).
2) Shorten y to two seconds of sound (# samples = 2 sec * 8000 samples/sec).
3) Use a switch statement that branches according to the string variable called effect as follows.
4) If the value of effect is 'clip', set all sound samples greater than 0.1 to 0.1 and samples less than –0.1 to –0.1, and play the sound. (This operation clips off peaks.)
5) If the value of effect is 'squish', delete all sound samples that are greater than 0.2 or less than –0.2, and play the sound. Use a single instruction for this step. If you try to use a for loop, the indexing into the y array will change every time a sample is deleted, causing the bookkeeping to become a nightmare.
6) If the value of effect is 'rectify', set all sound samples less than 0 to 0, and play the sound.
7) If the value of effect is 'sine_mask', set samples to the larger (in magnitude) of two waveforms: the sound in y, and a 100 Hz sinusoid of magnitude 0.2.
8) If the value of effect is not recognized as one of the above terms, display the message "unsupported effect".
Test each effect and use the listen command to play the sounds.
C. Sound Effect based on Median Filter
Create a script file called medfilter.m that uses a for loop to step through a sound snippet already stored in y and turn it into a stair-step waveform. The medfilter script operates as follows:
1) At the command prompt (rather than in the script file), create a variable called nsamps that is equal to the last digit of your student ID number, unless the last digit is less than two. If the last digit is less than two, use ten plus the last digit of your student ID number. The value of nsamps specifies the block size for processing samples of the sound. The value of nsamps also specifies the length of the stair-steps in final waveform. Your script file will assume the value of nsamps is already defined when the script file is run.
2) In medfilter.m, start by checking for nsamp < 2, which is an error. If this error occurs, medfilter.m should display an appropriate message and return.
3) After error checking in medfilter.m, use a for loop to extract the next nsamps of the sound waveform in y each time through the loop. An easy way to do this is to use the colon operator to create the for loop indexing variable as an array from 1 to the length of y and incremented by nsamps. The indexing variable will then contain the starting address in y of the current block being processed.
4) Replace the current block in y with the mean value (i.e., average value) of the current block in y. That is, replace every value in the current block with the average value of the block. This process will cause the final waveform to have stair-steps that are nsamps long.
Note: it may be helpful to use concatenation such as yout = [yout, ymed]; to add samples at the end of an output array each time through the for loop. In this case, ymed would be nsamps of computed in step (4), above. At the end of the script file, you can set y equal to yout. Note that this approach requires a statement such as yout = [] before starting the for loop.
D. Sound Effect based on Repeated Samples
Create a function file (not a script file) called rep.m that has the following input variables:
% invec input sound array
% start_samp sample number where repeated sound will start
% rep_size size in samples of repeated sample
rep.m will output a variable called outvec containing the sound array in invec with a section of repeated samples inserted in it. That is, the function should extract a section of invec of lenght rep_size and then insert a copy of that extracted section at the end of the section. This will make the sound longer, and the result will be a reverberation or stuttering effect. Experiment with rep.m to find an interesting value of rep_size. Play the sound for your TA.
E. Sound Effect based on Echoes
Create a function file (not a script file) called echoes.m that has the following input variables:
% invec input sound array
% echo_delay number of samples after which echo starts
% echo_size loudness of echo relative to original (1.0 means equally loud)
echoes.m will output a variable called outvec containing the sound array with echoes in it. The output sound is constructed by adding copies of the original sound scaled by echo_size and delayed by echo_delay samples. Recall that 8000 samples equals one second. To avoid creating an overly loud sound, scale the final output sound so the median value is the same as it was before the echoes were added. That is, multiply the entire waveform by a constant so that the average or mean value of the waveform is the same as it was before echoes were added.
Each successive echo will be created by multiplying the previous echoe by echo_size. Thus, each echo is smaller than the last, and the delay for the echo increases by echo_delay. Note that the length of the final sound should be the same as the original sound. Eventually, the echoes will be dealyed so much that they no longer overlap the original sound and may be ignored.
F. Sound Effect based on Spectrum
Create a function file (not a script file) called spechange.m that has the following input variable:
% invec input sound array
spechange.m will output a variable called outvec containing the input sound array with an altered spectrum. Your function must first compute the fft (Fast Fourier Transform) or spectrum of the input sound array. Then your function will alter the spectrum in some way. The spectrum presents the sound in terms of its frequency content. You may think of the spectrum as phasors represnting the amount of each frequency in the signal from 0 to one-half the sampling rate of 8000 Hz. The first half of the spectrum computed by the fft command contains all the frequency information for the signal. The second half is redundant; it is equal to complex conjugate of the first half in reverse order. Altering samples at the beginning and end of the spectrum affects the low frequencies in the sound. Altering samples in the middle of the spectrum affects the high frequencies in the sound.
The spectrum has complex values that are much like phasors. They encode the magnitudes and phase shifts of the sinusoids in the sound. They are in the rectangular format a + j*b, however. You can convert to magnitude and phase by using sqrt and atan2. This may be helpful if you decide to process the spectrum based on magnitudes or phase shifts.
After you have altered the spectrum, use ifft to convert the spectrum back into a time-domain waveform. Since the fft is complex and your alterations to the sound's spectrum may not result in ifft being real, you will probably need to apply the real function to your final sound array. Note: when plotting complex values, you can apply the abs function to extract magnitude.
Be creative in the ways you alter the spectrum of the sound. You may alter the spectrum in any way you like as you try to create an interesting result.
G. Original Sound Effect
Create a function file (not a script file) called effect.m that implements a unique sound effect of your own design. Your sound effect function may have whatever arguments you wish after the first argument, invec, which is the initial sound:
% invec input sound array
effect.m will output a variable called outvec containing the input sound with your sound effect applied to it. Note that your function must work with any input sound. Include comments in your file to describe the purpose of each line of code.
You will likely finish your sound effect outside the lab period. When it is finished, send it to your TA in whatever format the TA specifies.
Ref: [1] The Mathworks, Inc, Matlab® Primer, Natick, MA: The Mathworks, Inc, 2012.