Multi-harmonic Analysis Using Magnitude-Squared Coherence and Its Application to Detection of Auditory Steady-State Responses

Abstract

The detection of auditory evoked brain responses is an important task in hearing science, especially in the role of investigation of hearing thresholds. Objective Response Detection (ORD) techniques aim to identify the presence of evoked potentials based purely on statistical principles that perform an automatic hypothesis test in the frequency domain and the Magnitude-Squared Coherence (MSC) is a well-known and very efficient uni-variate ORD technique. The use of q-sample tests, which, in addition to the fundamental frequency, also includes higher harmonics in the detection has shown trends to better detection of ASSRs performance. The database used in this work contains ASSRs that were collected when evoked by amplitude modulation of pure tones delivered binaurally at 70 dB SPL to 24 volunteers with normal hearing thresholds. This paper analyses the detection of response using a multi-harmonic approach combining the fundamental frequencies, 84 and 88 Hz, and its six next harmonic frequencies. A detection threshold was estimated using a Monte Carlo simulation. Both the detection rate and area bellow the detection curve increased using q-MSC techniques when compared to the one-channel and one-harmonic technique. The best results trends to be using a mean value (mean q-MSC) up to the third harmonic frequency, with an increase of 7.4% of detection rate mean, statistically proven with McNemar test, and the mean area bellows the detection curve increased 24.37%, statistically proven with t paired test, for the 14 channels compared.