TY - JOUR
T1 - Automatic training of phoneme dictionary based on mutual information criterion
AU - Okawa, Shigeki
AU - Kobayashi, Tetsunori
AU - Shirai, Katsuhiko
N1 - Funding Information:
The antliors wonld like 1.0 tliank Akitiiro Yaiiai for helping with the experiments. This work is partly siipport,ed by tlie Grant-in-Aid for Scieii tihc Research from the Ministry Education, Science and Culture of Japan, No. 05241107 REFERENCES [l] Special Issue on Speech Database. J. Acoirstrcol .Sacit t y of Jnpnn. Vol. 48, No. I?.D FC 1992. [?] Y. Gong and I . PI. Halon: D'T\C'-Rascd Phonetic I a- iug Explicit l'lioncme Diiratioii Const.rainLs, 'LP 99. Fr.d.tv:2..3. Oct.. 199'2. [:iS]. Kapadia. \:. Va1t.chr.v and I;. J. Youiig: MMI Train-ing for Continuow Phoneme Recognition 011 the TIMlT Ilat.abase. Pi.ot. Il'.IS.YP-Y.3. II-491. Mar. I!l93. [4] li. Shirai. S. Okawa and T. Iiobayaslii: Phoneme Recvg-nitioii in Cont,iiiiious Speecli Based on Miit.ual Informa- tion Considering Plioneinic Lhratioii and Connectivity, !'roc. ICSLP 92. Fr.hhl.P,O ct.. 1Y92.
Publisher Copyright:
© 1994 IEEE.
PY - 1994
Y1 - 1994
N2 - Proposes an automatic training mechanism for phoneme recognition using labelless speech data under the condition that only its orthographical phonemic symbol sequence is given. For the purpose of obtaining better recognition performance the authors attempt to realize an automatic labeling procedure based on a phoneme classification method by mutual information criterion. By iterative training of a phoneme dictionary for a large amount of speech data, one can investigate the performance and convergence properties of the dictionary. From experimental results, the percent correct of the labeling is over 98% after three iterations, and for the phoneme recognition, a very high accuracy is also obtained.
AB - Proposes an automatic training mechanism for phoneme recognition using labelless speech data under the condition that only its orthographical phonemic symbol sequence is given. For the purpose of obtaining better recognition performance the authors attempt to realize an automatic labeling procedure based on a phoneme classification method by mutual information criterion. By iterative training of a phoneme dictionary for a large amount of speech data, one can investigate the performance and convergence properties of the dictionary. From experimental results, the percent correct of the labeling is over 98% after three iterations, and for the phoneme recognition, a very high accuracy is also obtained.
UR - http://www.scopus.com/inward/record.url?scp=85031633295&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85031633295&partnerID=8YFLogxK
U2 - 10.1109/ICASSP.1994.389310
DO - 10.1109/ICASSP.1994.389310
M3 - Conference article
AN - SCOPUS:85031633295
SN - 0736-7791
VL - 1
SP - I241-I244
JO - Proceedings - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing
JF - Proceedings - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing
M1 - 389310
T2 - Proceedings of the 1994 IEEE International Conference on Acoustics, Speech and Signal Processing. Part 2 (of 6)
Y2 - 19 April 1994 through 22 April 1994
ER -