Improving speaker identification robustness to highly channel-degraded speech through multiple system fusion | IEEE Conference Publication | IEEE Xplore