Analog Classification of Human Hand Gestures Exploiting Dendritic Integration With an Organic Artificial Soma

ABSTRACT An organic neuromorphic architecture for signal classification that generates a discriminative analog output from muscular activity is demonstrated. It operates as a dendritic integrator, formed by an artificial postsynaptic soma with three dendrites (a branched poly(3,4‐ethylenedioxythiophene)/polystyrene sulfonate ‐ PEDOT/PSS ‐ microelectrode), each receiving input from an artificial presynaptic neuron (a PEDOT/PSS microelectrode in a solid electrolyte). Each presynaptic neuron receives the input signal from one of three muscles. The architecture performs a classification task after proper adjustment of the synaptic weights, expressed in terms of solid electrolyte concentrations. The output from the artificial soma, expressed in terms of exchanged charge, is shown to clearly differentiate between the three archetypal gestures “rock”, “scissors”, and “paper” from the traditional “Ro‐sham‐bo” game. These results demonstrate that a passive organic neuromorphic architecture based on dendritic integration can directly process and discriminate raw bioelectrical signals at the hardware level through solid‐electrolyte‐defined synaptic weighting. This proof‐of‐concept highlights the potential of organic neuromorphic electronics for future low‐power bioelectronic interfaces and localized signal processing.