Audio mining is usually used in speech recognition software and music analysis. This technology gives the user the ability to search through speech or music audio that has been analyzed for specific characteristics. When used in speech recognition technology, audio mining identifies spoken words in the audio and puts them in a searchable file. This feature can be useful for students or those in the business world who attend many meetings because it allows the user to more easily browse topical information from speech presentations. This type of analysis can also be used in music to determine characteristics like beats per minute (BPM), musical key, and musical structure, information that is employed to classify music.
In speech recognition, where the technology is most often used, audio mining is employed to create an acoustic model. An acoustic model programs speech recognition software to recognize speech patterns as words. This technology is developed by audio mining a recording of a spoken phrase, which is compared to text matching the spoken phrase. The computer uses the information to recognize words when the user makes similar sounds to those in the acoustic model. An acoustic model is used in combination with a file that tells the speech recognition program what language to interpret and what patterns of words are likely to be spoken in certain sentences and situations.
Musicians and music listeners can both benefit from audio mining in music. Sometimes, music software that categorizes music by genre uses audio mining to organize the music. The process identifies and groups music files with sound similarities that frequently occur in musical genres. Though this technology can make organizing music and finding new music easier, it can make mistakes classifying songs that have similar measured characteristics but different overall sound. Audio analyzing software can be useful to musicians, especially composers, because it allows the composer to jump to specific parts of the song structure, including musical key changes and words within the lyrics.
The speech recognition software manufacturer Dragon® sells a program called AudioMining® that transcribes audio files and marks the files so they can be searched for text. Dragon is a manufacturer of computer linguistics programs, the technical term for the field of software designed to interpret speech. Audio mining, when used as two words, is a general term that refers to analyzing a sound file for a determined set of audio characteristics. Other manufacturers of audio mining software include Nuance® and Nexidia®.
