Evidence of stochastic resonance of auditory steady-state response in electroencephalogram for brain machine interface

Stochastic resonance is a phenomenon observed in nonlinear systems for which random noise with optimal level amplifies a weakly periodic signal. In some biological systems, stochastic resonance has been found to be utilized to improve signal transmission. Recently stochastic resonance have been evidenced in photic-driven human electroencephalogram (EEG) and demonstrated to improve performance of brain machine interface (BMI) based on steady state visual evoked potentials. The present study is aimed at giving evidence of stochastic resonance behavior in human auditory steady state response (ASSR) in EEG for developing a high-performance auditory BMI available without visual function. Seven healthy subjects aged 21-24 years old with normal hearing ability participated in the experiment in which their EEG responses to sinusoidally modulated tone with modulation frequency of 40 Hz contaminated by random noise were measured over the entire scalp with varying the carrier frequency (500 and 4,000 Hz), sound pressure of the tone (40-60 dB) and the random noise level (0-50 dB). In four subjects, ASSR amplitude showed a bell-shaped fluctuation with a maximum at noise level of 40 or 50 dB following an increase of noise level, hence the stochastic resonance effect may be elicited in the auditory system. Moreover in the four subjects, we investigated the times when ASSR significantly appeared under two conditions of no noise and the optimal noise that maximized ASSR amplitude. With addition of optimal noise, detection time of ASSR was shortened in three subjects, and ASSR was elicited in other subject. Detection time of ASSR at optimal noise was distributed between three and seven seconds across subjects. These results will be necessary in order to design novel ASSR-based BMIs. Further investigation on the stochastic resonance behavior would provide useful observation for development of auditory BMIs with high classification accuracy by improving the signal to noise ratio in the modulation of ASSR associated with user's intent.