Method for pitch-synchronous speech parameterization with applications
- Summary
- Speech parameterization is used in speech recognition and synthesis to convert between audible speech and a manipulatable digital representation. However, current methods to parametrize voice signals can be corrupted by variations in pitch, leading to inaccurate speech recognition and robotic-sounding speech generation. This technology is a speech parameterization method that is able to completely separate and accurately quantify unique elements of human speech such as pitch and timbre, enabling more accurate speech recognition and more realistic-sounding speech generation.
- Technology Benefits
- More accurate modeling of human speech for speech recognition and speech synthesisCan be used with tonal languages (e.g. Mandarin Chinese)Compatible with any speech database used in traditional parameterization Can achieve higher voice quality than traditional speech coding methods at same bit-ratePatent Information:Patent Issued (US8,744,854)Patent Issued (US8,719,030)Patent Issued (US8,886,539)Patent Issued (US8,942,977)Patent Issued (US9,135,923)Tech Ventures Reference: IR CU16133
- Technology Application
- Speech recognition (speech-to-text)Speech synthesis (text-to-speech)Voice transformationSpeech coding
- Detailed Technology Description
- None
- *Abstract
-
None
- *Inquiry
- Bonnie GurryColumbia Technology VenturesTel: (212) 854-8444Email: TechTransfer@columbia.edu
- *IR
- cu16133
- *Principal Investigator
-
- *Publications
- Chen CJ. "Elements of Human Voice" World Scientific Publishing. 2016. ISBN 978-9881-4733-89-2
- Country/Region
- USA
For more information, please click Here
