Speaker Recognition with Two-Step Multi-Modal Deep Cleansing | IEEE Conference Publication | IEEE Xplore