Method and apparatus for segmenting a multimedia program based upon audio events
View Patent ↗The present invention provides for a method and apparatus for segmenting a multi-media program based upon audio events. In an embodiment a method of classifying an audio stream is provided. This method includes receiving an audio stream. Sampling the audio stream at a predetermined rate and then combining a predetermined number of samples into a clip. A plurality of features are then determined for the clip and are analyzed using a linear approximation algorithm. The clip is then characterized based upon the results of the analysis conducted with the linear approximation algorithm.
1. A method of classifying a segment of a multimedia program as one of a news segment and a commercial segment, comprising:
(a) receiving an audio portion of the multimedia segment;
(b) sampling the audio portion at a predetermined rate;
(c) combining a predetermined number of samples into a clip;
(d) determining a plurality of signal property features of the clip, the signal property features including a smooth pitch ratio;
(e) analyzing the plurality of signal property features of the clip using a linear approximation algorithm, the linear approximation algorithm being a hard threshold classifier algorithm wherein it is determined whether each of the plurality of signal property features of the clip satisfies a respective threshold value; and
(f) characterizing the clip as the commercial segment if all of the plurality of signal property features of the clip satisfy the respective threshold value, and characterizing the multimedia program segment as the news segment if at least one of the plurality of signal property features of the clip does not satisfy the respective threshold value.
2. The method of claim 1 wherein the multimedia program is a television news program.
3. The method of claim 1 wherein the clip is comprised of a plurality of frames.
4. The method of claim 3 , wherein each frame of the plurality of frames overlaps at least one adjacent frame.
5. A process for separating commercials from television news programs comprising:
(a) receiving a television news broadcast containing an audio signal;
(b) extracting said the audio signal from the television news broadcast;
(c) sampling the audio signal at a continuous predetermined rate;
(d) grouping the audio samples together into an individual clip;
(e) analyzing of at least nine features of a first individual clip using a linear approximation algorithm, the features including a smooth pitch ratio, the linear approximation algorithm being a hard threshold classifier algorithm wherein it is determined whether each of the at least nine signal property features of the first individual clip satisfies a respective one of the at least nine threshold values; and
(f) characterizing the first individual clip as a commercial if all of the at least nine signal property features of the clip satisfy the respective threshold value, and characterizing the first individual clip as a news segment if at least one of the at least nine signal property features of the clip does not satisfy the respective threshold value.
6. The process of claim 5 wherein the hard threshold classifier has a smoothing algorithm.
7. The process of claim 5 wherein (f) further comprises:
(i) characterizing a second individual clip as a commercial;
(ii) if the first individual clip is characterized as a commercial, then characterizing a third individual clip between the first and second clips as a commercial.
8. The process of claim 5 wherein the continuous predetermined sampling rate is 16 kHz.
9. The method of claim 5 wherein grouping the audio samples together into an individual clip further includes first combining the samples into overlapping frames.
10. A method of classifying a segment of a multimedia program, comprising:
(a) receiving an audio portion of the multimedia segment;
(b) sampling the audio portion at a predetermined rate;
(c) combining a predetermined number of samples into a clip;
(d) determining a plurality of signal property features of the clip, the signal property features including a non-pitch ratio;
(e) analyzing the plurality of signal property features of the clip using a linear approximation algorithm, the linear approximation algorithm being a hard threshold classifier algorithm wherein it is determined whether each of the plurality of signal property features of the clip satisfies a respective threshold value; and
(f) characterizing the clip as a commercial segment if all of the plurality of signal property features of the clip satisfy the respective threshold value, and characterizing the multimedia program segment as a news segment if at least one of the plurality of signal property features of the clip does not satisfy the respective threshold value.