Kardome expands speech recognition enhancement portfolio with voice biometrics
Israeli startup Kardome has added voice biometrics and wake-up word detection to its speech recognition product enhancement portfolio.
The first of the two features is designed to enable developers and original equipment makers to build voice interfaces that recognize individuals and respond only to them.
“Our voice biometric technology works with embedded systems, offering secure and fast response times, and does not require Internet connectivity,” Kardome executives have written on their website.
Furthermore, the company says it offers voice profiles that can be securely stored online so people speaking can be identified across different systems regardless of location.
The wake word detection feature reportedly relies on Kardome’s deep learning and noise reduction algorithms to provide highly accurate detection and integrate with the company’s echo-cancellation and noise-reduction products.
According to Kardome, its wake word technology offers a false response rate of less than 10 percent with a signal-to-noise ratio of -20db. The software is compatible with Qualcomm, HIFI4, ARM ADSP-SC58x/59x devices and others.
More generally, Kardome’s 3D Audio Front End software is designed to enhance speech signals from multiple speakers and in real time.
According to CEO Dani Cherkassky, adding wake words and voice biometrics will give developers of speech recognition systems a one-stop solution for their voice technology.
“Our goal is to help developers and manufacturers of voice-controlled devices to create [automatic speech recognition] systems that are accurate, fast and secure,” Cherkassky says.
Kardome’s expanding voice recognition product line parallels developments in voice biometrics generally. In July, Nikolay Gaubitch, Pindrop’s director of research, told Biometric Update how the firm intends to use its technology to update banking.