Room identification using acoustic features in a recording
Analysis of the audio component in multimedia data is disclosed. Rooms can be described through room impulse responses (RIR), the “fingerprint” of a specific room. The method uses machine learning techniques to identify rooms from ordinary audio recordings.
1. A method of identifying a room comprising the steps of:
storing a database of a plurality of acoustic models for a plurality of identified rooms;
inputting audio data from at least one microphone in an unknown room;
extracting Mel-Frequency Cepstral Coefficient audio features from the input audio data;
using machine learning to create a set of a plurality of parameters of functions as an acoustic model of the unknown room based upon the extracted Mel-Frequency Cepstral Coefficient audio features;
comparing the acoustic model of the unknown room to the plurality of acoustic models in the database to determine a likelihood score that the model of the unknown room compares to respective ones of the acoustic models;
determining the highest likelihood score of the unknown room to at least one of the acoustic models in the database;
identifying the acoustic model of the unknown room as similar to the room in the database based on the likelihood score of the unknown room; and
outputting the identity of the unknown room.
2. A method, of identifying a room comprising the steps of:
storing a database of a plurality of acoustic models for a plurality of identified rooms;
inputting audio data from at least one microphone in an unknown room;
extracting audio features from the input audio data;
using machine learning to create a set of a plurality of parameters of functions as an acoustic model of the unknown room based upon the extracted audio features, wherein the acoustic model is a Gaussian mixture model;
comparing the Gaussian mixture acoustic model of the unknown room to the plurality of acoustic models in the database to determine a likelihood score that the model of the unknown room compares to respective ones of the acoustic models;
determining the highest likelihood score of the unknown room to at least one of the acoustic models in the database;
identifying the Gaussian mixture acoustic model of the unknown room as similar to the room in the database based on the likelihood score of the unknown room; and
outputting the identity of the unknown room.
3. The method according to claim 1 or 2 , wherein the microphone is provided in a mobile device.
4. The method according to claim 3 , wherein the mobile device is one of a telephone, PDA, cell phone, camera and hearing aid.
5. The method according to claim 3 , further comprising the step of determining a location of the mobile device using one of WiFi and GPS.
6. The method according to claim 1 or 2 , wherein if no match is determined, output a result that the unknown room is not one of the rooms in the database.
7. The method according to claim 1 or 2 , wherein the set of the plurality of parameters of functions created by machine learning include a plurality of curves.