Speech Recognition

  • HOME > 
  • Speech Recognition > 
  • Selvy STT-Server

Selvy STT-Server

stt system
Selvy STT has exclusively developed core elements in the overall speech recognition (STT server) process, from speech detection, feature extraction and sound effect modeling to navigation and post-processing. We have secured a high recognition of the Korean language by reflecting the language’s characteristics and vocalization habits, not to mention the support of high-quality consultancy through our R&D and commercialization performance over the last 10 years.
  • Expands/modifies the scope of words recognized with features such as isolated command, large-scale POI, variant words and connective words
  • Features various recognition engines applicable to H/W specifications and application scenarios
  • Allows the speaker to repeat utterances and prevents incorrect recognition by assessing the user ¡s vocalization history
  • Ensures recognition performance in diverse, noisy environments by adopting multiple sound effect modeling tools and multi-phase recognition methods
  • Enables an efficient and shortened development period by configuring a concise API
  • Requires only a small amount of system resources to run
  • Supports a convenient upgrade feature by modulating the recognition engine library
  • Strong noise-proof recognition performance is achieved by applying multi-channel noise reduction technology
  • Solution proven through obtaining the Good Software (GS) certification

Selvy WiseTell

speech to text software
WiseTell is an interactive speech recognition solution based on conjunctive keywords for the establishment of a speech recognition call-center. With the aid of continuous speech recognition (STT server) technology in a call-center, a customer can make an inquiry in more natural speech rather than in short questions, increasing both the customer satisfaction and the processing speed.
Major Application Examples  
ARS at call-centers
Applied to voice services searching for variable customer information in the ARS (Banking, securities, insurance, cards, transportation, aviation, weather, culture, customer center)
Voice Portal
Speech recognition technology (STT server) used to search for information on
websites and telephone networks in real time using voice.
Speech Recognition Automated Attendant (AA)
Speech recognition technology (STT server) used by an unmanned telephone automated
attendant (AA) system to search for a department or a person to call using voice.
 

Features and benefits of the solution

speech recognition ars
Product
specifications
Selvy STT-Server
Resources ROM : 200MB
RAM : 500 MB by default, 20 MB/CH
(based on 100,000 words)
Standard H/W
specifications
2.0GHz Xeon * 2EA (8 core) CPU 4GHz, DL380 G6
Channel capacity Maximum 60 channels
(based on 100,000 words)
Response speed 0.69 seconds (average of 60 channels)
OS MS Windows (2000, XP, Server)
Linux Red Hat (7.X, 8.0, 9.0)
Items Server-Client Stand-Alone
Sampling 8kHz DSR(8kHz,16kHz) 8kHz 16kHz
Speaker-
independent
Yes Yes Yes Yes
Number of
words recognized
Around 100,000
Capable of
recognizing
connective words
Yes Yes Yes Yes
Capable of
recognizing
variant words
Yes Yes Yes Yes
Response speed 2 seconds 2 seconds 2 seconds 2 seconds
Applied field CTI, VAD Mobile, Potal,
Robot
CTI Kiosk, Robot