SpeechPro Releases Version 1.2 of VoiceGrid SDK

SpeechPro has released VoiceGrid SDK version 1.2, the newest addition in its line of biometric voice identification tools. The software development kit (SDK) can be used to develop unimodal and multimodal biometric systems to perform accurate and rapid voice identification.

The VoiceGrid SDK 1.2 improvements include the following:

• A diarization algorithm: two separate voice models can be extracted and built from one recording featuring 2-speaker dialog;
• Greater search speed: VoiceGrid SDK 1.2 can perform between 2,000 and 32,000 comparisons per second on a single core processor (depending on the algorithms applied);
• 64-bit compatibility: version 1.2 works with 64-bit Windows out-of-the-box, facilitating work with in-memory databases containing millions of voice samples;
• Additional sample quality data: all quality estimation data, including pure speech duration, SNR, and reverberation time, is now available via API;
• "Hot Lists" functionality: users can now create in-memory lists with code names and perform 1:N searches against those lists;
• Deeper load testing: the SDK has been tested and approved for use on databases of up to 2 million voice models; and
• Optional Linux version: available upon request.

"The fact that VoiceGrid SDK 1.2 has a significantly wider functionality set and offers improved efficiency on standard SpeechPro features will be of great benefit to our partners who are developing voice identification applications for their customers," said Alexey Khitrov, president of SpeechPro. "In terms of search speed, version 1.2 is significantly faster, taking less than a minute to search and match through 1 million voice models. The automatic two-person dialogue segmentation feature, also called diarization, is a major plus when it comes to voice sample gathering and creating voice models."