Yoshiaki Itoh

J-GLOBAL         Last updated: Apr 22, 2014 at 16:51
 
Avatar
Name
Yoshiaki Itoh
E-mail
y-itohiwate-pu.ac.jp
Affiliation
Iwate Prefectural University
Section
Faculty of Software and Information Science

Research Areas

 
 

Misc

 
Yoshiaki Itoh, Hiroyuki Saito, Kazuyo Tanaka, Shi-Wook Lee
IPSJ Journal   54(12) 2492-2501   Dec 2013
We propose a method based on pre-retrieval results using syllable N-grams as query terms in advance in the Spoken Term Detection (STD). In the proposed method, all the combinations of syllable N-grams such as syllable bigram and trigram are search...
Jinki Takahashi, Kazuma Konno, Masumi Kumagai, Shi-wook Lee, Kazuyo Tanaka, Kazunori Kojima, Masaaki Ishigame, Yoshiaki Itoh
IPSJ SIG Notes   2013(3) 1-7   Dec 2013
We propose various methods for Spoken Term Detection (STD), which identifies the target utterances where query terms are spoken in spoken documents. In this paper, we apply three methods to STD to improve the retrieval accuracy. The experimental r...
Yoshiaki Itoh, Hiromitsu Nishizaki, Seiichi Nakagawa, Tomoyosi Akiba, Tatsuya Kawahara, Xinhui Hu, Hiroaki Nanjo, Tomoko Matsui, Yoichi Yamashita, Kiyoaki Aikawa
IPSJ Journal   54(2) 471-483   Feb 2013
Spoken Document Retrieval (SDR) was dealt with in one of tracks of TREC from 1996 to 2000. NIST supplied a task for Spoken Term Detection (STD) in 2006. Many researches have been conducted as for SDR and STD after these projects. A working group f...
KAMATA KEISUKE, SAITO HIROYUKI, ITOH YOSHIAKI, KOJIMA KAZUNORI, ISHIGAME MASAAKI, TANAKA KAZUYO, LEE SHI-WOOK
IEICE technical report. Speech   112(369) 137-142   Dec 2012
We have been conducting a research for Spoken Term Detection (STD), which identifies the target section where query terms are spoken in spoken documents. In STD, Out-Of-Vocabulary (OOV) query terms are one of the most important problems because OO...
Nakazawa Shogo, Mikawa Masahiko, Tanaka Kazuyo, Itoh Yoshiaki
Technical report of IEICE. PRMU   112(357) 1-6   Dec 2012
This paper proposes an improved method for real-time music piece identification in a cover performance stream. The previous method performed with low accuracy because of susceptibility to acoustic feature variations. In addition, it performed with...