Sound Search Results

The new YouTube feature (currently implemented on a small scale) of searching for videos based on content within it adds a new dimension to searching on the Web by finding even more meaningful information. Although it is not something pioneered or perfected by Google, it brings speech recognition to the limelight.

A few other interesting applications of speech recognition for public consumption that I have come across are -

Blinkx - a video search engine has 26 million hours of video.

Airtel, a popular mobile service provider in India, has a feature called 'SongCatcher' that lets you hold your mobile against any television, music system or radio playing your favorite song for 20 seconds & sets that as a "caller tune".

The Music Recognition Technology that enables this quick and easy access to music is based on an algorithm that is capable of identifying patterns in music by matching voice frequencies. The ‘DNA’ of the requested song is matched with the 25,000 songs available on Hello Tunes! and is accurate to the point of distinguishing between the music and the chatting of a DJ or a reporter.

The Microsoft Research Speech Group has developed speech recognition features like Speech Indexing and Query-By-Humming and Music Steering-

Enabling computers to be smart about speech and audio is a primary focus of the Speech Group. A core innovation that has come out of this is a search engine that can index the words spoken in recorded conversations whether they are from meetings, conference calls, voice mails, presentations, online lectures, or even video. Microsoft OneNote 2007, a part of Microsoft Office, is the first Microsoft product to include our speech-indexing technology to allow users to search for keywords spoken in recorded meetings and phone calls.

We have teamed up with Windows Live Mobile China to build a prototype service - call the service, hum a tune, get your favorite ringtones. "Music steering" means interactive music playlist generation through music content analysis, music recommendation, and music filtering. With a personal music collection of thousands of songs in our portable devices such as iPhone, Zune, and smart phones, selecting songs has become a challenge. Music steering provides a “smart shuffle” function: Pick a starting (seed) song, and the system will automatically build a playlist of similar songs. It can be refined by voting up/down and setting a “mood filter.” At the foundation is technology to analyze music content and automatically detect musical attributes (tags) from each song like genre, instruments, tonality, and tempo/rhythm.