Analysis
Analysis of
of Sonification
Sonification for
for EEG
EEG
Classification
Classification
ISTEC General Assembly 2014
STUDENTS: STUDENTS:
GONZÁLEZ GONZÁLEZ CASTAÑEDA ERICK CASTAÑEDA ERICK FERNANDOFERNANDO
MScMSc. TORRES GARCÍA ALEJANDRO ANTONIO. TORRES GARCÍA ALEJANDRO ANTONIO
ADVISORS: ADVISORS:
DR. VILLASEÑOR PINEDA LUISDR. VILLASEÑOR PINEDA LUIS
Introduction
2
Nearly 200 million people experience some difficulties in motor skills.
Brain Computer Interfaces (BCI) are an alternative for integrating this people using BCIs and provide them some autonomy.
Problematic
3
The electrophysiological sources like SCP, P300, VEP, Motor imagery, have a long training period and low rates of communication.
Exploring the use of Unspoken Speech (or Imagined Speech) as source for BCIs.
EEG Sonification
4
Sonification (or Auditory display) refers to the use of non-speech audio to transmit information.
Objective
5
Methodology
6
Complete Methodology
Experimentation and Results
7 66,06 38,18 39,39 46,67 61,82 31,52 48,78 66,06 58,79 41,21 73,33 49,70 49,09 32,73 59,39 37,58 49,70 58,79 27,88 75,76 31,52 55,15 47,27 33,33 28,48 41,21 49,39 70,30 40,00 55,76 58,18 75,15 30,91 52,44 73,33 60,61 48,48 83,03 59,39 60,61 41,82 64,85 46,06 55,76 54,55 32,12 73,33 49,70 68,48 57,58 60,00 33,94 44,85 56,10 0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 S22 S23 S24 S25 S26 S27
Accuracy results using 4 channels and Random Forest
Conclusions
8
Sonification was applied to transform the EEG signal to an
audio signal.
The classification accuracy was improved, by choosing
dominant spectrogram frequencies from EEG signal and mapping EEG frequencies to audio frequencies.
In future work, we will explore using 14 channels and
additional feature extraction methods such as MFCC.
48,10
55,83
0,00 10,00 20,00 30,00 40,00 50,00 60,00
Average
Average accueracy using 4 channels and RF