Although we did not create MP3 files, we used the concepts that we learned in ELEC 301 to understand how MP3 works. We were excited to learn that the digital signal processing concepts we learned in class have such useful real-world applications. By doing this project, we looked into great depths of how data compression, audio signals in particular, is achieved. We were able to produce audio files that are much smaller in size but still retain decent sound quality. In particular, the Psychoacoustic Model works the best in terms of the amount of compression and the sound quality of the output.
Future Improvements:
If we had more time, we would try
to implement good windowing and temporal masking. We did not try to implement
any lossless compression in this project. The reason being is that lossless
compression really does not have anything to do with DSP. Lossless compression
would be making a .zip file, which is really comp. sci. And, of course, a real
sound compression program should be real-time when it comes to decompression.
Real-time decompression might impose extra constraints on the types of algorithms
that can be used in compression, but real-time stuff is outside the scope of
this project.