Using low precision floating point numbers to reduce memory cost for MP3 decoding | IEEE Conference Publication | IEEE Xplore