Smart Equalizer - Artificial Neural Networks in Music
Grigoruta Alexandru
coord. Lect. dr. Ignat Anca
Table of Contents
1. Presenting the hypothesis
2. Sound equalization
3. Proposed architecture
4. Results
5. Conclusions
Presenting the hypothesis



Shuffled Playlist
Manual Equalization
Just Enjoy
Automatic Equalization
Photo by Alexey Ruban on Unsplash
Photo by Patrik Michalicka on Unsplash
Photo by Kelvin Lutan on Unsplash
Sound equalization

Sound equalization - Parameters
Cutoff Frequencies

Central Frequency
Bandwidth
Gain
Sound equalization - Types

Shelving Equalizer

Graphic Equalizer
Parametric Equalizer


Implementation steps
1. Research similar papers or applications
2. Research audio signal processing papers
3. Understand current situation and progress in the ASP community
4. Define and design multiple architectures
5. Implement and optimize the most viable architecture
6. Train the architecture on a big dataset
7. Test the agent with different audio files, optimize based on findings
Proposed architecture

Proposed architecture - Genre Detection



Proposed architecture - Automated EQ












Results
Pink Floyd - Wish You Were Here
Equalized
Original
Wolfgang Amadeus Mozart - Eine Kleine Nachtmusik KV 525
ZZ Top - La Grange
Conclusions
Optimizations
Future Work
Different Architecture
Bigger, more diverse dataset
Target gains set by a number of professional sound engineers
User Interface
Learning user behaviour and equalizing based on it
Integration with existing or new music streaming service
Thank you!
Smart Equalizer - Artificial Neural Networks in Music Grigoruta Alexandru coord. Lect. dr. Ignat Anca
Smart Equalizer - Artificial Neural Networks in Music
By alexgrigi
Smart Equalizer - Artificial Neural Networks in Music
- 633