Microsoft Word - Assignment2.doc
1. Answers to questions with appropriate predictions, calculations,
MATLAB graphs and discussions.
2. References: Any sources used to find out more information. I.e.
textbooks, journal/conference papers, websites, etc. must be in
IEEE style format.
3. Appendix: MATLAB code and other results and calculations.
The assignment will be assessed on the depth of conceptual understanding
shown for each task. It is important not only to present co
ect results/
graphs/code but also to be able to analyse and discuss what your results are
showing and how they link in with the concepts behind FIR and IIR filters.
Assignment 2: FIR Filters in Audio Processing
This assignment will focus on the filtering of audio signals using FIR filters (specifically the
audio file ‘countdownfrom.wav’ available to download from Canvas). Audio signal
processing uses many types of digital filters to create effects such as echoes (often using FIR
filters) and reve
erations (using IIR filters), or to remove unwanted artefacts in a signal (e.g.
noise or channel effects).
This assignment will be looking at how we can simulate a concert hall type effect on audio
using an FIR filter.
a) In Audio editing software packages often contain effects to allow you to simulate
different recording environments. An example could be to simulate the audio being
produced in a concert hall, auditorium or pipe by adding echoes to the audio or to
simulate a padded room by filtering the audio to dampen the sound.
In this task you are required to simulate the echo-ey room shown in figure 1:
Figure 1: An echo-y sound environment
Description:
- This room contains a speaker in the top right-hand corner of the room which
produces sound for the listener shown in the centre of the room.
- Our listener experiences three versions of the audio;
- The direct version coming straight from the speaker (blue line). There is no
attenuation or delay with this audio (apart from the time it takes to get to the
listener).
- A first echo bouncing from the right-hand wall (red line) which is delayed by
Td1 seconds (relative to the direct path), the amplitude of this audio has also
een attenuated by G1 of the direct audio’s amplitude and,
- A second echo bouncing from the left-hand wall (green line) which is delayed
y Td2 seconds (relative to the direct path), the amplitude of this audio has
also been attenuated by G2 of the direct audio’s amplitude.
The delays and attenuations will be based on the 5th number in your student number:
Student
Number
Delays Attenuations
1 Td1 = 0.15
Td2 = 0.2
G1 = 0.8
G2 = 0.4
2 Td1 = 0.2 Td2
= 0.25
G1 = 0.7
G2 = 0.35
3 Td1 = 0.25
Td2 = 0.3
G1 = 0.7
G2 = 0.25
4 Td1 = 0.3 Td2
= 0.35
G1 = 0.6
G2 = 0.2
5 Td1 = 0.35
Td2 = 0.4
G1 = 0.4
G2 = 0.1
6 Td1 = 0.35
Td2 = 0.45
G1 = 0.3
G2 = 0.2
7 Td1 = 0.3
Td2 = 0.5
G1 = 0.2 G2
= 0.15
8 Td1 = 0.25
Td2 = 0.55
G1 = 0.2
G2 = 0.1
9 Td1 = 0.2
Td2 = 0.6
G1 = 0.2
G2 = 0.1
0 Td1 = 0.1
Td2 = 0.2
G1 = 0.9 G2
= 0.45
Table 1: Delays and attenuations on sound paths
- For example if your student number was XXXXXXXXXXyour first path will be
delayed by 0.2 sec and attenuated to 20% of the original amplitude, your
second path will be delayed by 0.6 sec and attenuated to 10% of the original
amplitude.
i. Manually work out (neatly typed) the number of samples your delays will
co
espond to and the weightings (ak) for each part of the filter.
ii. Manually work out (neatly typed) the difference equation, impulse response and
transfer function for this filter.
iii. Show the Direct Form I diagram for this filter (use a software package such as
Microsoft Visio to show this neatly).
iv. Now use MATLAB to import the speech file ‘countdownfrom.wav’ which you will
need to download from Canvas. Using the filter coefficients and delays you manually
calculated above, create the FIR filter and use either the conv()or filter()
functions to filter the ‘countdownfrom.wav’ file.
Listen to this new audio file; you should hear an echo effect occu
ing now.
Discuss:
1. Explain in your report how this filter generates the echoes. What’s happening?
2. Plot the time-domain waveform for your filtered and non-filtered waveform,
identify where the two echo paths are showing in the signal?