NSA’s Method for Recognizing Sounds of Speech of Any Language

Agency: 
Dept. of Defense

Laboratory: National Security Agency

Technology: Method of Recognizing Phones in Speech of Any Language

Opportunity: This unique NSA-patented technology is available for licensing by interested parties.

Details: This technology processes spoken sounds, independent of the language used, and translates them into text. Using a unique, universal alphabet comprised of phonetic sounds from approximately 15 languages, this technology enables the ability to train speech recognition systems to recognize sounds from different languages.

Benefits: Speech processing applications rely on their corpora for translation—that is, their internal database of text that aids in translation. Multilingual speech processing applications typically require corpora from each language to be merged into one large corpora. This is problematic as the applications can sometimes produce inaccurate or unintelligible results. This technology’s built-in corpora, or database of writings, allows speech processing applications to identify sounds from multiple languages without the need for manual checking or outside databases.

Applications:

  • Language training and evaluation of artificial intelligence
  • Foreign language and English as a Second Language (ESL) courseware development
  • Medical and legal transcription and translation
  • Media and broadcasting
  • Government, military, and law enforcement applications

Contact: For more information about this NSA technology, contact Linda Burger at tech_transfer@nsa.gov.

To view this and thousands of other available technologies in FLC Business, visit flcbusiness.org.

Region: 
Mid-Atlantic