DSP Project Proposal
DSP Project Proposal
1. Project Goal 3
2. Project Description 3
3. Implementation plan 6
3.1. Timeline 6
4. Summary 6
Project Goal
The project aims to demonstrate practical application of different types of filters in de-noising
audio signals. We wish to investigate appropriate filter design to eliminate noise patterns in a set
of different noise-corrupted audio signals.
Project Description
We have a noisy audio signal. The original speech and the Signal to Noise ratio are both
unknown. Our method de-noise the signal is to extract information of noise patterns from the
corrupted audio signal by analyzing the power spectral density of the signal. Based on the
observations in the spectral density, appropriate filter type and filter parameters would be
determined. Next, the filters would be applied to the noisy signal and filter performance would
be noted. Filter coefficients would be modified until satisfactory levels of signal to noise ratio are
achieved. This process would be repeated for a set of noisy signals to experience application of
different filter designs.
Data collection
Noisy audio file samples can be found on several websites the internet. One such website is
http://signalogic.com/index.pl?page=codec_samples.
MATLAB functions/tools
There would be a main function which would read the file, cut the file into frames, plot
power spectral density function of the frames, apply filter functions to the frames, output
the filtered frame to an audio file for performance check and lastly repeat the process for
the entire signal.
MATLAB ‘FDAtool’
The project features design of different filters for the same/different audio signals. This
includes changing filter type and specifications. The use of FDAtool would be preferred
in this project for its convenience and quick design benefits over command line filter
design.
The “Filter Design and Analysis Tool (FDATool)” is a graphical user interface (GUI) in
MATLAB Signal Processing Toolbox™. With this, we can quickly design FIR/IIR filters
by setting up requisite filter specifications. The FDAtool can be called up by the ‘fdatool’
command.
Filter functions would be designed by using the FDAtool. Filter parameters would be set
up after observing the power spectrum of the noisy audio signal. Using FDAtool, these
filter parameters could be easily adjusted as well.
The tentative flow of the program, source of data, MATLAB tools used and MATLAB function
to be built are mentioned in the coming sections.
The project can be distributed into nine tasks. These tasks and the milestones have been listed in
the time schedule below. Working period for the project would be 15th October, 2013 to 11th
December, 2013. This period has eight weeks listed as week-1(Wk-1) to week-8(Wk-8) in the
time frame below. After the initial 2 weeks of project planning and study, both of us would work
Timeline
Task Wk-1 Wk-2 Wk-3 Wk-4 Wk-5 Wk-6 Wk-7 Wk-8
Project Planning
Collection of Data
Review filter types/parameters
MATLAB coding
Spectrum analysis
Filter design
Results/Evaluation
Report Writing
Oral presentation/Demo
Summary
Filter design is an elementary application in digital signal processing. Establishing this project
would be a good learning experience in filter design and application. We think that the project
demonstration at the end of the class could serve as a good tutorial for the students as well. Using
this experience, we can explore advanced topics in signal processing in future.