1Pocketsphinx 0.8 2^^^^^^^^^^^^^^^^ 3 4Thanks: Tanel Alumae, Douglas Bagnall, Halle Winkler, Peter Grasch, Vyacheslav Klimkov 5Erik Andresen, Nicola Murino, Melmahdy, Scott Silliman, Riccardo Magliocchetti, Marc Legendre, 6Kho-And-Mica, Zheng6822, Pankaj Pailwar, Evandro Gouvea 7 8Improvements: 9 * API is better exposed in Python, SWIG and gstreamer 10 * New API for exact grammar match 11 * Configurable type for the frame counter, allows utterances way longer than 5 minutes 12 13Fixes: 14 * Memory leaks, refcounting and other memory-related issues 15 * Use proper word bounary senones for the words added on the fly 16 * Accurate FSG lextree construction 17 18Pocketsphinx 0.7 19^^^^^^^^^^^^^^^^ 20 21Thanks: David Huggins-Daines, Chen Tao, creative64, Edwin Miguel, Bhiksha Raj 22 23Improvements: 24 25 * Nice performance improvements 26 * Use arc iterators to build FSG lextrees, speeds this up immensely 27 * Filler word propagation is responsible for 5% speedup 28 * Continuous decoder can decode files, not from the microphone 29 * Add senone score dumping and reloading capability 30 * Functional alignment pass 31 * Symbian support 32 * Add the Android makefile and other JNI stuff 33 * Case sensitive names, including phone names 34 * Better error messages 35 * Report timing for individual search passes 36 37Bug fixes: 38 39 * Accuracy regression fixes 40 * Fixes the bug with reading transition matrix on big-endian platforms 41 * Fixes very important regression with NULL transitions in fsg search 42 * Proper acoustic score scaling during posterior calculation. 43 44And many, many, many more intersting things! 45 46Pocketsphinx pre 47^^^^^^^^^^^^^^^^ 48 492000-01-27 CMU Sphinx is going Open Source! 50