Emotional feedback for service robots using hapto-acoustic interface

A behavior-oriented human-machine communication system based on bio-inspired information processing is developed in this paper. A low cost acoustic interface using microphones is created as an emotional interaction channel for robotic assistants in domestic environments for the care of elderly people with mild cognitive impairment. Classification of communication signals such as striking, petting, speech and noise could be successfully realized using 3 condenser microphones and deployed on a robot prototype. The proposed framework comprised of an electronic microphone interface circuit, an ATmega128 microcontroller for processing of signals from the circuit, a Fast Fourier Transform (FFT) algorithm implementation on the microcontroller for Fourier analysis, LUFA library for USB data transfer from the microcontroller, a machine learning module based on Support Vector Machine (SVM) for signal classification and an amplifier based on an RFT transistor capsule for feedback. Results show that in all four cases, a successful differentiation with reliable and robust deviation between the signals can be made. The success rate with the SVM classifier shows with 87.5 % a highly promising classification result. It can be concluded that this novel approach is highly useful for non-verbal human-machine communication.