|
|
Project Description
Modified Goals and Objectives
In our implementation we used filter banks, psychoacoustic analysis, and
different quantization schemes to perform compression on audio signals.
First, our code takes an audio signal in .wav file format and runs it through
a set of 32 filter banks, which divides the signal into its different frequency
components. While the signal is being divided by the filter bank, psychoacoustic
analysis is performed on it as well. The psychoacoustic model tells us
how much noise can be added to each frequency range of the filter bank.
With this information, we can determine the amount of quantization error,
and hence the number of bits, to use for each set of frequencies that make
up the signal. This then allows us to compress the audio signal and
get an approximate size of the output file. We also use different quantization
schemes to vary the amount of compression we get on the signal.
Finally, we then reconstruct the signal and through auditory testing and frequency
domain analysis determine the quality of the compressed signal.
Procedure:
- research filter bank methods that allow for even breakup and high-quality reconstruction of a signal
- implement filter bank scheme
- research psychoacoustic model
- implement psychoacoustic model which takes signal and returns allowable noise over frequency range
- quantize signal based on amount of allowable noise
- determine amount of space necessary to save the signal using this method and compare to saving without compression.
|
