Feature discovery using snap-drift neural networks

This paper introduces an application of Snap-Drift Neural Networks (SDNNs), which employs the complementary concepts of fast, minimalist (snap) learning and slow (drift towards the input pattern) learning, for feature discovery and classification of speech waveforms from nonstammering and stammering speakers. The speech waveforms are drawn from a phonetically annotated corpus, which facilitates phonetic interpretation of the classes of patterns discovered by the SDNN. The results show that SDNN groups the phonetics speech input patterns meaningfully and extracts properties which are common to both non-stammering and stammering speech, as well as distinct features that are common within each of the utterance groups, thus supporting classification. SDNN is also being applied in a virtual learning environment to categorise students’ test responses and thereby support individualised feedback.

[1]  Sin Wee Lee,et al.  Reinforced snap-drift learning for proxylet selection in active computer networks , 2004, 2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541).

[2]  Matthew P. Aylett,et al.  Vowel quality in spontaneous speech: what makes a good vowel? , 1998, ICSLP.

[3]  Dominic Palmer-Brown,et al.  Performance-guided neural network for rapidly self-organising active network management , 2004, Neurocomputing.

[4]  D H Klatt,et al.  Review of text-to-speech conversion for English. , 1987, The Journal of the Acoustical Society of America.

[5]  D. Palmer-Brown,et al.  Snap-drift: real-time, performance-guided learning , 2003, Proceedings of the International Joint Conference on Neural Networks, 2003..

[6]  Sin Wee Lee,et al.  Snap-drift learning for phrase recognition , 2005, Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005..

[7]  Dominic Palmer-Brown,et al.  The Analysis of Network Manager’s Behaviour using a Self-Organising Neural Networks , 2005 .

[8]  P. Ladefoged.  A course in phonetics , 1975 .

[9]  Teuvo Kohonen,et al.  Improved versions of learning vector quantization , 1990, 1990 IJCNN International Joint Conference on Neural Networks.