LumenVox Improves Core Speech Engine
LumenVox recently completed improvements to its core Speech Recognition Engine (SRE) v2.6, which also affects the Speech Driven Information System (SDIS).
These improvements are focused on functionality and user development of the core engine. The C++ API is now a lightweight wrapper class that can be altered by users. With that, a new C API has been added, in which these C functions are called through the C++ LVSpeechPort class. Other changes to the API include the allowance of an utterance to easily be decoded with many user-defined grammars. The number of user defined grammar sets has also been increased to 64 for each LVSpeechPort, up from the original four, with each grammar set allowing 2,000+ words/phrases. The Speech Recognition Engine (SRE) v2.6 is now completely thread-safe, which allows developers to not worry about system interruptions. The out-of-vocabulary filter is also functional, so no extra time is needed to decode an utterance when the filter is on. The SRE has improved decode accuracy and confidence measures, improving the overall recognition quality. The ability to see individual phonemes or phrases that a user spoke is now also available. Due to the fact that the Speech Driven Information System (SDIS) is an easy-to-use GUI, wrapped around LumenVox's Speech Recognition Engine all of the above improvements are included with v2.6 of the SDIS.