My name is Aliaa Mahgoub. I am a junior at Brooklyn Technical High School.
Coming soon! Please refer to my project Google Site for the time being.
I experimented with the number of mel bins in the mel spectrogram to see how it affects the different evaluation metrics using both the microphone array (MIC) and first-order Ambisonics (FOA) formats. The original number of mel bins used in the model is 64.
As expected, the FOA data resulted in better scores overall. However, the FOA and MIC formats had different optimal nmel numbers. The MIC data performed best overall with 256 Mel bins, while the FOA data performed best with 128 Mel bins.
GitHub: AliaaMahgoub
Email: aliaamahgoub0@gmail.com
LinkedIn: Aliaa Mahgoub