Digital Signal processing: IIR
filters
This assignment covers IIR filters. Your task is to solve a problem which requires realtime filtering of a physical
quantity. You have one option for the data acquisition: • If you don't want to spend any money you can use your webcam as an RGB colour
sensor, sampling at its frame rate. Every team needs to have a different application for their measurement. Think of
something simple such as measuring the speed of a fan with an LDR or the webcam. Perhaps you have a fishtank and you'd like to know how often a fish swims past. Or if the
flowerpot is watered. Add your topic to the wiki provided on moodle as soon as possible. It needs to solve a practical real life problem and require low, high, bandpass or bandstop
filtering. Excessive use of identical topics will result in low marks. Again you work in teams of two students and one report is submitted per team. 1. Present a realtime measurement problem to be solved which requires filtering. For
example noise needs to be removed or a signal needs to be detected. Marks are
given for initiative, inventiveness and originality (= ideas which haven't come
straight from the lecturer, lab demonstrators or other groups). Document the
experiment with (all compulsory):
[30%]
• photos of the setup
• dataflow diagrams
• YouTube clip(s)
in addition to your report. How would you like to present the results? Just as a plot
or perhaps a bar graph? QT for Python might be an option to look into. [20%]
2. Check the sampling rate of your acquisition by running it for example
for 10secs and check against the number of samples expected. How could you
check for jitter in the sampling? [20%]
3. Determine the filter response(s) which are required and justify them. Generate the SOS coefficients for the filter(s) either with the help of Python's high
level functions or analytical solutions shown in the lecture. [20%]
4. Write two classes:
(a) IIR2Filter which implements a 2nd order IIR filter which takes the
coefficients in the constructor and has a method called:
y=IIR2Filter.filter(x)
where y and x are simple scalars (no arrays) as usual. Optimise this class that
it won't need any arrays for its buffers and coefficients. (b) a class IIRFilter which directly takes the SOS array from the high level IIR
design commands as its constructor argument and which then creates a chain
of 2nd order filter instances of IIR2Filter classes. Thus they form an array of
instances of IIR2Filter. Again implement a function which then filters the
signal:
y=IIRFilter.filter(x)
and then internally processes the data x by sending it through the chain of 2nd
order IIR2Filter classes. Implement the filtering operation again in the most
effective way by not using index operations. 5. Compare your filtered results with the original recordings, show both signals in a
realtime demo (YouTube clip) and discuss if you have been successful. Do a
critical analysis. [10%]
High level design commands are allowed such as “butter” but the actual IIR filtering
operations need be written from scratch as outlined above. Any use of “lfilter”，“conv” or
other high level python filter operation will result again in zero marks. Proof of realtime
processing in form of a video needs to be given and the video needs to show clearly what
it's about. Please add your link to the wiki and also add a readme to your zip containing all
the files. The code needs to be again submitted as a zip file and it will be tested if it runs. Crashing code will result in low marks. As before I expect sharp figures in vector format in the report. The complete code
needs to be in the appendix and also uploaded to moodle. Deadline for the report is 2nd Dec.