Using the Fisher kernel method for Web audio classification Export

@inproceedings{16243, abstract = {As the multimedia content of the Web increases techniques to automatically classify this content become more important. We present a system to classify audio files collected from the Web. The system classifies any audio file as belonging to one of three categories: speech, music and other. To classify the audio files, we use the technique of Fisher kernels. The technique as proposed by Jaakkola (1998) assumes a probabilistic generative model for the data, in our case a Gaussian mixture model. Then a discriminative classifier uses the GMM as an intermediate step to produce appropriate feature vectors. Support vector machines are our choice of discriminative classifier. We present classification results on a collection of more than 173 hours of Web audio randomly collected. We believe our results represent one of the first realistic studies of audio classification performance on found data. Our final system yielded a classification rate of 81.8\%}, author = {Moreno, P. J. and Rifkin, R.}, citeulike-article-id = {3355144}, citeulike-linkout-0 = {http://dx.doi.org/10.1109/ICASSP.2000.859329}, doi = {10.1109/ICASSP.2000.859329}, journal = {Acoustics, Speech, and Signal Processing, 2000. ICASSP '00. Proceedings. 2000 IEEE International Conference on}, keywords = {audio, automata, classification, classifier, content, databases, discriminative, feature, fisher, gaussian, generative, information, internet, kernel, learning, library, machines, method, mixture, model, multimedia, music, probabilistic, probability, resources, speech, support, systems, vector, vectors, web}, pages = {2417-2420 vol.4}, posted-at = {2008-09-30 05:56:26}, priority = {2}, title = {Using the Fisher kernel method for Web audio classification}, url = {http://dx.doi.org/10.1109/ICASSP.2000.859329}, volume = {6}, year = {2000} }

TY - CONF ID - 16243 L3 - citeulike-article-id:3355144 N2 - As the multimedia content of the Web increases techniques to automatically classify this content become more important. We present a system to classify audio files collected from the Web. The system classifies any audio file as belonging to one of three categories: speech, music and other. To classify the audio files, we use the technique of Fisher kernels. The technique as proposed by Jaakkola (1998) assumes a probabilistic generative model for the data, in our case a Gaussian mixture model. Then a discriminative classifier uses the GMM as an intermediate step to produce appropriate feature vectors. Support vector machines are our choice of discriminative classifier. We present classification results on a collection of more than 173 hours of Web audio randomly collected. We believe our results represent one of the first realistic studies of audio classification performance on found data. Our final system yielded a classification rate of 81.8% JF - Acoustics, Speech, and Signal Processing, 2000. ICASSP '00. Proceedings. 2000 IEEE International Conference on EP - 2417-2420 vol.4 TI - Using the Fisher kernel method for Web audio classification VL - 6 SP - 2417-2420 vol.4 KW - audio KW - automata KW - classification KW - classifier KW - content KW - databases KW - discriminative KW - feature KW - fisher KW - gaussian KW - generative KW - information KW - internet KW - kernel KW - learning KW - library KW - machines KW - method KW - mixture KW - model KW - multimedia KW - music KW - probabilistic KW - probability KW - resources KW - speech KW - support KW - systems KW - vector KW - vectors KW - web AU - Moreno, PJ AU - Rifkin, R PY - 2000/// UR - http://dx.doi.org/10.1109/ICASSP.2000.859329 ER -