Hidden Markov models with spectral features for 2D shape recognition

Cai, Jinhai; Liu, Zhiqiang

doi:10.1109/34.977569

Cited by 33 publications

(1 citation statement)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…EEG also provides an interesting means to understand brain functionality during haptic perception [9]. Researchers have also performed object shape recognition from tactile images through various methods like the use of neural networks [10], [11], regional descriptors [12], image gradient [13], fuzzy classification and reconstruction of 2-D shapes [14], use of Markov models for 2-D shape recognition [15] to mention a few.…”

Section: Introductionmentioning

confidence: 99%

Correlation analysis of object shape recognition from EEG and tactile signals

Khasnobish

Datta

Pal

et al. 2014

2014 International Conference on Advances in Electrical Engineering (ICAEE)

View full text Add to dashboard Cite

Touch perception can be decoded from brain signals as well as from tactile pressure signals obtained while holding objects using a sensor. The present work aims to classify object shapes from EEG signals and tactile pressure signals obtained from a capacitive tactile sensor while exploring objects of different shapes; thereby drawing a mutual relation or dependence between these two sources of haptic information in response to the same tactile stimuli. It is evident from the classification results that object-shapes can be classified efficiently from both EEG and tactile signals with mean classification accuracy of 74.21% and 97.12% respectively. Correlation analysis using various metrics show that EEG and tactile signals are non-linearly correlated and only a very small amount of non linear correlation exists. The polynomial fitting of EEG signals on tactile signals depicts that in absence of brain signals tactile signals can successfully predict the corresponding EEG signals.

show abstract