New capabilities to compliment voice biometrics announced by Sensory, AWS, LumenVox
Sensory has announced support for multiple wake words among a set of new capabilities for its TrulyHandsfree offering, enabling custom branded wake words, small vocabulary command sets, and even user-defined wake words from any combination of leading brands, according to a company announcement.
The technology supports the efforts of the Voice Interoperability Initiative (VII), which was announced in September, and allows companies to begin developing products featuring multiple simultaneous voice services.
Sensory says it has already developed deep learning models for “Hey Cortana”, “Alexa”, “OK Google”, “Hey Siri”, “Hi Bixby”, “Xiao-du Xiao-du” and more than two dozen other assistants. The above wake words are approved for the market and shipping in products with proven performance and reliability in ultra-low power applications like wearables and smartphones, as well as vehicles, smart speakers and home appliances, according to the announcement.
“We are lucky to have worked closely with most of the leaders in voice services to help create products with their wake words embedded. Because of this, we already have all of the data in place to combine multiple models,” says Todd Mozer, CEO of Sensory. “However, everyone in the industry has found that combining two or more wake word models into a product poses a significant challenge in maintaining optimal performance. The new capabilities we added to TrulyHandsfree overcomes some of the accuracy challenges without significantly boosting MIPS and memory requirements.”
The new technique for multiple wake words employed by TrulyHandsfree avoids linearly increasing error rates, which has been the main challenge for running multiple wake words on one device, the company says. Unlike traditional approaches, Sensory’s technology does not compound the error rates of each individual wake word, resulting in less false acceptances and less false rejections.
Mozer says the company expects to release a multi-wake word adaptation that allows devices to biometrically identify the speaker and adapt the voice model accordingly during 2020.
Sensory will demonstrate TrulyHandsfree by appointment at CES 2020.
Amazon launches medical transcription service
The new Amazon Transcribe Medical was announced at AWS re:Invent, and does not require doctors to say things like ‘comma’ to insert punctuation in the text. The text can also be provided to other systems or services like Amazon Comprehend Medical. The company says the service is HIPAA-eligible and offered on a scaling per-use basis.
Collected PCM audio is sent to a streaming API based on the WebSocket protocol, and the API returns JSON blobs of transcribed text with time stamps, which can optionally be saved to an Amazon S3 bucket.
LumenVox voice technology compliant with Avaya Contact Center offerings
LumenVox has announced that a series of its voice solutions are now compliant with Avaya IX Contact Center solutions.
The LumenVox Speech Automation Suite has been compliance-tested by Avaya, and integrates with the Avaya Aura Experience Portal 7.2.2 to deliver accurate, natural sounding speech for improved customer experiences.
“Our Speech Automation Suite has been available through the Select Product Program for several years, providing LumenVox partners and customers with an easier ordering process and competitive price point,” comments LumenVox CEO Edward Miller. “This latest testing with the Avaya Aura Experience Portal demonstrates our continued commitment to delivering reliable, natural sounding speech technology, helping our mutual customers improve customer service experiences.”
LumenVox is a Technology Partner in Avaya’s DevConnect program, through which the compliance and compatibility were assured. The company also updated its authentication suite with new biometric features earlier this year.