VoiceIn 4.1 product specification sheet
New Fonix VoiceIn Features
- Phonetic localization:
- Improved phoneme estimates for better recognition and language acquisition applications
- Improved rejection of non-speech sounds
- Improved speech detection and speech framing
- Improved demo applications
- New platforms: Blackfin, Java-WinCE
Accurate
- Noise-tolerant
- Insensitive to environment acoustics (no training necessary)
- Performs with background speech and music
- Multi-pass recognition
- Optimized letter (spelling) and digit recognition (voice dialing)
Efficient
- Compact neural networks/small memory footprint
- Small computing requirements (starts at 20 MIPS)
- Dynamic optimization, minimize memory and MIPS
- Recognition time – 100 ms after end-of-speech detection
- Proprietary optimized memory manager for fileless OS support
Flexible
- Speaker-independent, continuous speech
- Word spotting / Finite-state grammars
- Dynamic word spotter and grammar creation during run-time
- Create grammars/words lists from unlimited vocabulary
- N-best support for word spotting, grammars, and spelling
- Adjustable out-of-vocabulary rejection
- Normalized confidence scoring for phonemes, words or phrases
- Speaker pronunciation modeling
- Speaker dependent phonetic voice tags
- Far-field microphone capabilities
- Offline grammar and runtime compilers
- User data association with recognition words
- Wide character support
Vocabulary
- Active vocabulary size is only limited by memory and MIPS
- Customizable user dictionary
- Supports spelling greater than 65 K words
Languages
- UK/US English, Canadian/European French, German, Italian, Japanese, Castilian/Latin American Spanish, Swedish, Korean
- Mandarin Chinese (in development)