New focus assessment method for iris recognition systems

Iris Image Blur Detection with Multiple Kernel Learning

“…The results show a significant improvement in performance afforded by merging the ESDD and SCH features over using them separately. Moreover, ESDD+SCH performs significantly better than the algorithms proposed by Ma [6] and Jang [7]. The aver- age cost time of ESSD+SCH, and the methods proposed by Ma [6], and Jang [7] are: 0.14 s, 0.04 s, and 0.07 s on synthetic dataset and 0.14 s, 0.04 s and 0.06 s on real dataset, respectively.…”

“…To further evaluate the utility of our proposed algorithm, it was compared against two other methods [6], [7]. [6] uses the frequency spectrum energy of high-, middleand low-frequency as blur feature and [7] uses the ratio of the average wavelet coefficients in high and low frequency bands.…”

“…[6] uses the frequency spectrum energy of high-, middleand low-frequency as blur feature and [7] uses the ratio of the average wavelet coefficients in high and low frequency bands. The final correct detection rate of every method and its standard deviation for 10 time tests on synthetic and real datasets are reported in Tables 1 and 2.…”

“…In addition to the above methods which define blur features on frequency spectrum, there also exist some approaches which model blur features according to wavelet coefficients. Jang et al [7] estimated the proportion of average wavelet coefficients in high and low frequency bands and defined this as blur feature. Chen et al [10] proposed to use Mexican hat wavelet to extract wavelet coefficients of local region at different scales as blur feature.…”

A Emergency Sound Detecting Method for Smarter City

“…As human speech generally falls within 300-3400khz, the FIR filter [12][13][14] can separate the incoming audio data into passband (the speech region), stopband, and threshold-band.…”