David Kryze Perrine Delacourt and Christian J. Wellekens. Speaker-based segmentation for audio data indexing. In SCA International Speech Communication Association, Cambridge, UK, pages 78--83, 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....window. By keeping the window size sufficiently large to reliably apply the method, and sufficiently short to avoid multiple transitions, a segmentation algorithm can be devised that relies on the basic q case. In Figure 1 an algorithm is proposed [5] that was derived by the one described in [7]. The main idea is to have a shifting variable size window in which a SC can be hypothesized according to (9) To reduce computations, the maximization (7) is not computed over all points 30 q , but at a lower resolution rate. The resolution rate is increased when a potential SC is ....

P. Delacourt, D. Kryze, and C. J. Wellekens. Speaker-based segmentation for audio data indexing. In Proceedings of the ESCA ETRW workshop Accessing Information in Spoken Audio, Cambridge, UK, 1999.

....and sufficiently short to avoid multiple transitions, a segmentation algorithm can be devised that relies on the basic c = 1 case. In Figure 1 an algorithm is proposed that exploits an adapting window size and resolution of the maximum search. The algorithm [5] is inspired by the one described in [7]. The main idea is to have a shifting variable size window in which a SCs can be hypothesized according to (9) To reduce computations, the maximization (7) is not computed over all points 1 : n 1, but at a lower resolution rate. The resolution rate is increased when a potential SC is ....

P. Delacourt, D. Kryze, and C. J. Wellekens. Speaker-based segmentation for audio data indexing. In Proceedings of the ESCA ETRW workshop Accessing Information in Spoken Audio, Cambridge, UK, 1999.

....adaptation, and 3.4 after adaptation. Introduction Automatic segmentation, classification and clustering in terms of acoustic source and acoustic channel of an audio stream is a challenging problem to which many research groups have recently devoted attention (Chen Gopalakrishnan, 1998; Delacourt et al. 1999; Gauvain et al. 1999; Hain et al. 1998; Harris et al. 1999; Liu Kubala, 1999; Tritschler Gopinath, 1999) A relevant application is, for example, the automatic indexing of multimedia digital libraries by using automatic speech recognition to transcribe audio signals containing speech. ....

....multiple changing points based on the BIC method. Since computation of BIC values is costly, several approximations of the BIC method and alternatives for detecting multiple changes were proposed, in order to make the computation more efficient but keeping the effectiveness of the approach (e.g. (Delacourt et al. 1999; Liu Kubala, 1999; Tritschler Gopinath, 1999) We implemented an algorithm, reported in Figure 3, inspired by that proposed in (Delacourt et al. 1999) The main idea is to have a shifting variable size window for the computation of BIC values. Moreover, in order to save computations, BIC ....

[Article contains additional citation context not shown here]

Delacourt, P., Kryze, D. & Wellekens, C. J. (1999). Speaker-based Segmentation for Audio Data Indexing. In Proc. of the ESCA ETRW workshop. Accessing Information in Spoken Audio, Mller centre, Cambridge, UK.

No context found.

P. Delacourt, D. Kryze, C.J. Wellekens, Speaker-Based Segmentation for Audio Data Indexing, ESCA-ETRW Workshop: Accessing Information in Spoken Audio, Cambridge (UK), April 1999.

....as significant when the differences between its value and those of the minima surrounding it are above a certain threshold (calculated as a fraction of the variance of the distance distribution) and when there is no higher local maximum in its vicinity. This detection method is detailed in [1]. Since a missed detection (an actual speaker turn has not been detected) is more severe for the verification process than a false alarm (a speaker turn has been detected although it does not exist) parameters involved in the speaker turn detection have been tuned to avoid missed detection to the ....

P. Delacourt, D. Kryze, C.J. Wellekens, Speakerbased segmentation for audio data indexing, ESCA workshop: accessing information in audio data, 1999.

....certain threshold (calculated as a fraction of the graph variance) and when there is no greater local maximum in its vicinity. Thus, the selection of the local maxima is not done considering the absolute value of the peaks, but rather by considering the form factor of the peaks, as detailed in [13]. Since a missed detection (an actual speaker change has not been detected) is more severe for the grouping process than a false alarm (a speaker change has been detected although it does not exist) parameters involved in the speaker change detection have been tuned to avoid missed detection to ....

....is discarded and then, both segments are merged to form one segment for the next pair of segments. 3. EXPERIMENTATIONS 3.1. Data Different types of speech data have been used to compare our segmentation technique with the algorithm proposed by S. Chen, referred to as the BIC procedure (see [3, 13]) ffl 2 conversations which are artificially created by concatenating sentences of 2 s on average from the TIMIT database (clean speech, short segments) ffl 2 conversations created by concatenating sentences of 1 to 3 s from a French language database provided by CNET (Centre National ....

[Article contains additional citation context not shown here]

P. Delacourt, D. Kryze, and C. J. Wellekens, "Speakerbased segmentation for audio data indexing," in ESCA workshop: accessing information in audio data, 1999.

No context found.

David Kryze Perrine Delacourt and Christian J. Wellekens. Speaker-based segmentation for audio data indexing. In SCA International Speech Communication Association, Cambridge, UK, pages 78--83, 1999.

No context found.

Delacourt, P., Kryze, D. &Wellekens, C.J. (1999). Speaker-based Segmentation for Audio Data Indexing. In Proceedings of EuroSpeech99 (pp. 1195- -1198).

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