|
While pursuing all major avenues of
speech technologies Speereo Software focuses on speech interfaces
based on speech commands. The underlying proprietary speech
recognition system is capable of performing tasks of high complexity.
It successfully solves problems currently encountered by many
developers.
Our speech recognition system offers a
number of unique advantages that can be easily revealed upon testing:
-
High accuracy of speech
recognition
For 600 command
phrases or phrases that can be recognized at the same time the
current accuracy is 99.9% in an office setting and 97% in a car
going at 60 mph with
windows down.
-
Extensive vocabulary
Vocabulary can be as big as to include thousands of commands and
multiple synonyms to allow for intuitive formulating of commands.
-
Short latency
Recognition
time for 1,000 words or phrases is 0.19 sec on PII450 processor.
-
Absolute speaker independence
User does not waste time and
does not experience discomfort adapting to the system or adjusting it to
his / her speaking manner, dialect or accent.
-
Semantic competence
Our
intelligent system understands the meaning of a command no matter
how verbally different it is expressed.
-
Ease of use
All one needs to
make our speech recognition system work is to start talking to it.
Speereo's speech recognition computer
program is written in C++ and has a simple API to interface with
other software or devices. This allows to perform speech
recognition on a variety of end-user devices having modest
processing capabilities (such as pocket-size computers, smart
phones, set-top boxes, or home appliances), as opposed to performing
it on servers. As a result, high recognition accuracy, short
latency, and vast vocabulary of speech commands are achieved.
Although
our speech engine supports only English at this time (performing well with
different accents), we are working on other major languages. We also
develop acoustic signal compression (50 times and more) algorithms
working in real time on processors of 40 Mips and above.
|
