Comparing neuromorphic encoding and analysis methods for texture classification using touch sensors

Kock, Robin (2025) Comparing neuromorphic encoding and analysis methods for texture classification using touch sensors. Master's Thesis / Essay, Artificial Intelligence.

Preview	Text mAI2025KockRT.pdf Download (3MB) | Preview
	Text Toestemming.pdf Restricted to Registered users only Download (159kB)

Abstract

Touch enables humans to effortlessly assess object properties, but robotic systems lack this ability. While advanced tactile sensors provide high-resolution pressure signals, traditional machine learning methods are computationally expensive. Neuromorphic computing, integrating memory and computation akin to the brain, presents a promising alternative. This study explores neuromorphic methods for classifying surface textures from pressure data. Sampled pressure data is converted into spike trains using biologically inspired encoders. These spike trains are analyzed using an array of spiking phase-locked loops (sPLLs) to extract frequency components, which are then classified using a linear regression model. The objective of our research is to investigate which neuromorphic encoding method yields the highest classification accuracy and whether a network of sPLLs can extract frequency components sufficiently to enable accurate classification. To this end, we optimize the sPLL network’s parameters for different encoders. The best network of sPLLs attained the highest accuracy of 95.4% using an encoder based on a leaky integrate-and-fire neuron. Generally, the sPLL network exhibited a performance comparable to that of a spiking fast Fourier transform, suggesting the potential of neuromorphic approaches in this domain. Furthermore, the findings of this study underscore the promise of neuromorphic methodologies in the analysis of touch sensor signals. Consequently, this contribut..

Item Type:	Thesis (Master's Thesis / Essay)
Supervisor name:	Chicca, E. and Besselink, B. and Pequeno Zurro, A. and Mastella, M.
Degree programme:	Artificial Intelligence
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	14 May 2025 13:10
Last Modified:	14 May 2025 13:10
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/35171

Actions (login required)

View Item