Yang L P, Young S T, Ku T S
Department of Electrical Engineering, National Taiwan University, Taipei, Republic of China.
Med Biol Eng Comput. 2002 Jan;40(1):34-40. doi: 10.1007/BF02347693.
Transiently evoked otoacoustic emissions (TEOAEs) are widely used in newborn hearing screening programmes for early detection of hearing losses. To increase the accuracy of a TEOAE pass/fail criterion that uses the wavelet method, it was demonstrated that the large estimation variance is a possible reason for the inaccuracy, and a modified wavelet method is proposed to solve the inaccuracy problem. In the modified wavelet method, N paired buffers, instead of only one, were used to store the total 512 subaveraged responses, and then the average of the calculated N cross-correlation coefficients between N pairs of TEOAE signals was taken in the pass/fail criterion. Four sets of 256 synthesised noise and eight sets of 256 synthesised noisy TEOAE signals were tested, and each set was tested 1,000 times. The results showed that the standard deviation of the correlation estimation was greatly reduced by using this average value with N selected as 4. As a result, the total number of single-scale cross-correlation coefficients below 50% decreased from 1281 to 195 for noisy TEOAE signals, and the total number of single-scale cross-correlation coefficients above 50% decreased from 90 to 0 for synthesised noise.
