References

Benchmarking

TravellingWave has conducted several internal experiments to compare VoicePredict against its voice recognition and text prediction counterparts. The tests were conducted as part of grant awards from the National Science Foundation. These tests were conducted in controlled environments, using a variety of different mobile handsets, and sampling several human users.

VoicePredict predicts words with 90% to 95% accuracy for just 1 to 2 letters input, which is a 60% improvement over competing voice recognition solutions. In comparison with state of the art word prediction solutions, VoicePredict results in an 80% reduction in key presses and a 25% increase in speed. Additionally, there is a 50 to 75% speed up in entry of contact-names, symbols, emoticons and sms lingo.

Publications

National Science Foundation Small Business Innovation Research Phase-I, Phase-IB, Phase-II, Phase-IIS grants; award number 0724271

"On Decomposing Speech into Modulated Components", Ashwin Rao and Ramdas Kumaresan, Journal of the IEEE Trans. On Speech and Audio Processing, May 2000

"Model Based Approach to Envelope and Positive Instantaneous Frequency Estimation of Signals", Ramdas Kumaresan and Ashwin Rao, Journal of the Acoustical Society of America, March 1999

Several patents issued and pending