Evidence for cognitive involvement in discriminating speech in babble.

INTRODUCTION Speech in Background babble Discrimination (SIBD) is a recognized challenge. Stochastic Resonance (SR) is offered as the cognitive mechanism for speech discrimination. OBJECTIVE We hypothesize that SIBD has the hallmarks of SR processing because it is a non-linear transformation that includes frequency spreading and allows for brain processing without a priori knowledge of the incoming signal. MATERIALS AND METHODS To test the hypothesis we created two sets of HINT based, Hagerman style vocabulary, tests. The stochastic signal used is a recording of speech induced emission noise. The first set determines the SR transformation settings using a multi-dimensional optimum search. The second set measures the actual improvement in speech recognition for normally hearing and hearing impaired subjects. RESULTS The tests show an improvement of approximately 2 to 3 dB SNR of speech intelligibility. The improvement is much more significant for Hearing Impaired (HI) subjects. CONCLUSION SR lowers the detection threshold of auditory signals by an order of magnitude. We show that these results, using emissions noise, show higher discrimination rate than reported results using non-matched noise. SR processing is a cognitive process as demonstrated by the sample sentences. OAE is suggested as the brain's mechanism of SR processing.