1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
|
Pocketsphinx 0.8
^^^^^^^^^^^^^^^^
Thanks: Tanel Alumae, Douglas Bagnall, Halle Winkler, Peter Grasch, Vyacheslav Klimkov
Erik Andresen, Nicola Murino, Melmahdy, Scott Silliman, Riccardo Magliocchetti, Marc Legendre,
Kho-And-Mica, Zheng6822, Pankaj Pailwar, Evandro Gouvea
Improvements:
* API is better exposed in Python, SWIG and gstreamer
* New API for exact grammar match
* Configurable type for the frame counter, allows utterances way longer than 5 minutes
Fixes:
* Memory leaks, refcounting and other memory-related issues
* Use proper word bounary senones for the words added on the fly
* Accurate FSG lextree construction
Pocketsphinx 0.7
^^^^^^^^^^^^^^^^
Thanks: David Huggins-Daines, Chen Tao, creative64, Edwin Miguel, Bhiksha Raj
Improvements:
* Nice performance improvements
* Use arc iterators to build FSG lextrees, speeds this up immensely
* Filler word propagation is responsible for 5% speedup
* Continuous decoder can decode files, not from the microphone
* Add senone score dumping and reloading capability
* Functional alignment pass
* Symbian support
* Add the Android makefile and other JNI stuff
* Case sensitive names, including phone names
* Better error messages
* Report timing for individual search passes
Bug fixes:
* Accuracy regression fixes
* Fixes the bug with reading transition matrix on big-endian platforms
* Fixes very important regression with NULL transitions in fsg search
* Proper acoustic score scaling during posterior calculation.
And many, many, many more intersting things!
Pocketsphinx pre
^^^^^^^^^^^^^^^^
2000-01-27 CMU Sphinx is going Open Source!
|
