Algorithm, technology being developed to improve speech recognition products

A approach to speech recognition technology? 

Waveform after years of Hidden Markov Model?

Published: June 5, 2012

INDIANAPOLIS – An Indiana startup is developing an algorithm that may improve voice searches on computers and smartphones, help people who wear hearing aids make a better distinction between similar sounding words, and has other applications.

Waveform Communication LLC, based in the Purdue Research Park, is developing digital code based on a model of vowel perception and production derived from cues in visual displays. The research was published in “The Waveform Model of Vowel Perception and Production,” by Michael A. Stokes, the company’s CEO.

Stokes said current speech recognition technology has flaws that limit its effectiveness.

“Some technology has only a 90 percent accuracy rate, which is unacceptable for the general public and professionals such as medical or court record transcriptionists,” he said. “Some technology requires users to train on their voice or cannot recognize multiple speakers.”

The Waveform Model achieves 99.2 percent voice recognition accuracy and can be used by multiple speakers without special training. Stokes’ code can be incorporated into electronic devices, and the company looks to develop a phone application within a year.

“Improved voice recognition systems in electronics may open up the world for people who cannot use or do not have access to a physical keyboard,” Stokes said. “Reducing errors in reports with highly technical language may lead to better end results for researchers and the general public.”

Waveform Communication secured $100,000 in funding from an Indiana-based venture capitalist in 2012. The company also was named one of the top 50 startups in the world by the Ewing Marion Kauffman Foundation in 2011.

Stokes explained why the company became a tenant in the Purdue Research Park of Indianapolis.

“The location by the Indianapolis International Airport is perfect for when my colleagues and partners need to travel to me,” he said. “The technology infrastructure at the park will accommodate the company’s growth, and the possibility of connecting with engineering resources from Purdue – including interns and Ph.D. consultations – will be invaluable.”

About Waveform Communication LLC

The Waveform Model is the first model that has demonstrated vowel perception at human performance levels. Beyond this, the Waveform Model also explains vowel production and perceptual errors. This is a complex cognitive process explained on the most fundamental level for the first time. Methods to improve speech in noise, speech recognition algorithm development, talker identification from waveforms, and improving hearing aids all are patent-pending. The successful innovation of the Waveform Model is leading to a number of application developments and is generating additional academic attention after the presentation of improved performance in 2011. The work will impact a number of technologies used every day and will continue to explain processes that have been studied for decades.

About Purdue Research Park

The Purdue Research Park, with four locations across Indiana, has the largest university-affiliated business incubation complex in the country. The parks are home to about 200 companies that employ 4,000 people and are located in West Lafayette, Indianapolis, Merrillville and New Albany.

Purdue Research Park contact: Steve Martin, 765-588-3342, sgmartin@prf.org

http://purdueresearchpark.com/node/1276

More-latest speech technologies
Social share or comment – what do you think?

There are no comments yet. Be the first and leave a response!

Leave a Reply

Wanting to leave an <em>phasis on your comment?

 
Trackback URL http://www.speechtechnologygroup.com/algorithm-technology-being-developed-to-improve-speech-recognition-products/trackback

Algorithm, technology being developed to improve speech recognition products

A approach to speech recognition technology? 

Waveform after years of Hidden Markov Model?

Published: June 5, 2012

INDIANAPOLIS – An Indiana startup is developing an algorithm that may improve voice searches on computers and smartphones, help people who wear hearing aids make a better distinction between similar sounding words, and has other applications.

Waveform Communication LLC, based in the Purdue Research Park, is developing digital code based on a model of vowel perception and production derived from cues in visual displays. The research was published in “The Waveform Model of Vowel Perception and Production,” by Michael A. Stokes, the company’s CEO.

Stokes said current speech recognition technology has flaws that limit its effectiveness.

“Some technology has only a 90 percent accuracy rate, which is unacceptable for the general public and professionals such as medical or court record transcriptionists,” he said. “Some technology requires users to train on their voice or cannot recognize multiple speakers.”

The Waveform Model achieves 99.2 percent voice recognition accuracy and can be used by multiple speakers without special training. Stokes’ code can be incorporated into electronic devices, and the company looks to develop a phone application within a year.

“Improved voice recognition systems in electronics may open up the world for people who cannot use or do not have access to a physical keyboard,” Stokes said. “Reducing errors in reports with highly technical language may lead to better end results for researchers and the general public.”

Waveform Communication secured $100,000 in funding from an Indiana-based venture capitalist in 2012. The company also was named one of the top 50 startups in the world by the Ewing Marion Kauffman Foundation in 2011.

Stokes explained why the company became a tenant in the Purdue Research Park of Indianapolis.

“The location by the Indianapolis International Airport is perfect for when my colleagues and partners need to travel to me,” he said. “The technology infrastructure at the park will accommodate the company’s growth, and the possibility of connecting with engineering resources from Purdue – including interns and Ph.D. consultations – will be invaluable.”

About Waveform Communication LLC

The Waveform Model is the first model that has demonstrated vowel perception at human performance levels. Beyond this, the Waveform Model also explains vowel production and perceptual errors. This is a complex cognitive process explained on the most fundamental level for the first time. Methods to improve speech in noise, speech recognition algorithm development, talker identification from waveforms, and improving hearing aids all are patent-pending. The successful innovation of the Waveform Model is leading to a number of application developments and is generating additional academic attention after the presentation of improved performance in 2011. The work will impact a number of technologies used every day and will continue to explain processes that have been studied for decades.

About Purdue Research Park

The Purdue Research Park, with four locations across Indiana, has the largest university-affiliated business incubation complex in the country. The parks are home to about 200 companies that employ 4,000 people and are located in West Lafayette, Indianapolis, Merrillville and New Albany.

Purdue Research Park contact: Steve Martin, 765-588-3342, sgmartin@prf.org

http://purdueresearchpark.com/node/1276

More-latest speech technologies
Social share or comment – what do you think?

There are no comments yet. Be the first and leave a response!

Leave a Reply

Wanting to leave an <em>phasis on your comment?

 
Trackback URL http://www.speechtechnologygroup.com/algorithm-technology-being-developed-to-improve-speech-recognition-products/trackback
css.php