B. Feiten and T. Ungvary, "Organizing sounds with neural nets," presented at the Proc. 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....of the audio les should be in terms of such classes of data for ecient retrieval. B. Categories of Audio Indexing tasks For ecient retrieval, some sort of audio similarity measure for organizing similar sounds into one category is desirable. Neural network based clustering of audio les [1] and self organizing maps (SOM) 2] have been used to organize similar sounds. Likewise at Muscle Fish (http: www.muscle sh.com ) Wold et al. have used a vector of attributes like loudness, pitch, bandwidth and harmonicity [3] to organize the sounds into similar classes. Such type of classi ....

.... where speech is rst converted to text prior to indexing, cepstral coecients have been the predominant choice as features for training Hidden Markov Models (HMMs) For the general case of audio indexing, features like energy contours, zero crossing rate and spectral centroid have been used [1]. Loudness, pitch, bandwidth and harmonicity have been used to classify music [11] Music speech discrimination has been done using the average length of time in which peaks exist in a narrow frequency range [12] To retrieve music, query by humming has been tried by Ghais et al. 5] In the case ....

B. Feiten and T. Ungvary, \Organizing sounds with neural nets," in Proc. Int. Computer Music Conf., 1991.

....or content. Rapid increase in the amount of audio data demands for a computerized method which allows efficient and automated content based classification and retrieval of audio database. Wold etal [17] have developed a system called Muscle Fish . That work distinguishes itself from earlier work [4, 5, 6] in its content based capability. There, various perceptual features are used to represent a sound. A normalized Euclidean (Mahalanobis) distance and the nearest neighbor (NN) rule are used to classify the query sound into one of the sound classes in the database. In Liu etal [10] separability of ....

B. Feiten and T. Ungvary, "Organizing sounds with neural nets", in Proceedings

....are crucial. While research in speech recognition, a closely related area, has a long history, research on content based classi cation and retrieval of audio sounds is relatively new. Foster et al. 7] aim to allow queries such as nd the rst occurrence of the note G sharp . Feiten and Ungvary [8] use a neural net to map sounds to text descriptions. Feiten and G unzel [9] use a self organizing map (SOM) to group similar sounds based on perceptually derived spectral features. An important recent work is done by Wold et al. 1] represented by their system called Muscle Fish . The work ....

....and G unzel [9] use a self organizing map (SOM) to group similar sounds based on perceptually derived spectral features. An important recent work is done by Wold et al. 1] represented by their system called Muscle Fish . The work distinguishes itself from the previous audio retrieval work [7] [8], 9] in its content based capability. In the Muscle Fish system, various perceptual features, such as loudness, brightness, pitch, timbre are used to represent a sound. A normalized Euclidean (Mahalanobis) distance and the nearest neighbor (NN) rule are used to classify the query sound into one ....

B. Feiten and T. Ungvary, \Organizing sounds with neural nets", in Proceedings 1991 International Computer Music Conference, San Francisco, 1991.

No context found.

B. Feiten and T. Ungvary, "Organizing sounds with neural nets," presented at the Proc. 1991.

No context found.

B. Feiten and T. Ungvary. "Organizing sounds with neural nets". In Proceedings

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC