Speech Recognition

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

Selvy STT-Embedded

As a speech recognition solution for devices, Selvy STT: Embedded guarantees efficient performance as optimized for environments with limited CPU or memory.
It provides stable recognition performance in noisy environments, such as in a vehicle, and is applicable to various types of mobile and digital devices(embedded stt).
Major Application Examples  
Extremely useful when setting the destination and navigating via voice
Black box/ Cell phone
Supports the "Voice Trigger" function to recognize the
device's(embedded stt) commands and execute the speech recognition actions.

Features and benefits of the solution


Implements a real-time recognition
feature for a large number words
in various types of hardware devices
and operating systems

Applies the existing speech recognition
method as well as notable phonemic
awareness-based recognition algorithms

Recognizes fixed vocabulary
as well as variable vocabulary

Utilizes a feature extraction
technique,reflecting the variation
of the input devices

Improves the recognition rate and
the reliability of the recognition results
by applying post-processing techniques
such as the models considering the
variation of pronunciation and
the surrounding noise

Equipped with a replaceable recognition
vocabulary set, which can be adapted for various application scenarios

Minimizes the memory usage
by compacting the millions of words
in the pronunciation dictionary

Applies sound model technology
which fully reflect Han-geul
phonetic values

Supports the Voice UI design,
optimized for the device environments
as well as the verification feature

User environment and product specifications

Items Selvy STT_Embedded
OS Windows CE, Embedded Linux, ISO, Android
CPU 300 MHz recommended for less than 1,000 words, and over 500 MHz recommended for around 100,000 words
RAM Requires 4 MB by default (Add up 1.2 MB per 100,000 words)
ROM Requires 2 MB by default (Add up 1.2 MB per 100,000 words)
Recognized words Up to one million