Deep bottleneck features and sound-dependent i-vectors for simultaneous recognition of speech and environmental sounds | IEEE Conference Publication | IEEE Xplore