著者

重本 賢太朗 清水 忠昭 鈴木 慶 吉村 宏紀 松村 寿枝

雑誌

情報処理学会論文誌 (ISSN:18827764)

巻号頁・発行日

vol.57, no.5, pp.1514-1523, 2016-05-15

---

我々は,空中手書き文字(AHC)入力システムのための自動的な文字分割手法の開発を行ってきた.これまでに提案した手法では,平仮名のAHCについて高い精度で文字分割に成功したが,連続して入力可能な文字数に制限があった.本稿では,同一領域に重ねて文字を入力する方法により入力文字数に制限のない文字分割手法を提案する.提案手法では,AHCのストロークを評価する5つのストローク評価指標を学習したサポートベクタマシン(SVM)によりストローク判別して文字分割を行う.提案手法によるストローク判別は,学習データでは,移動ストローク87.0%,文字ストローク96.5%の正解率となった.評価データに対しても,移動ストローク84.9%,文字ストローク96.1%の正解率を示した.試作した提案手法のデモ・システムについても紹介する.In this paper, we propose a segmentation method for an aerial handwriting character (AHC) input system. This work is an extension of a previously proposed hiragana AHC segmentation method that achieved high accuracy. However, its number of input characters was limited. In this paper, we propose a character segmentation method without such a limitation by overwriting characters on the same input area. Our method separates an AHC trajectory into characters by stroke distinction using a support vector machine (SVM) trained with five stroke evaluation indexes. The results of the evaluation experiments show that the detection accuracy was 80.9% for transition strokes and 98.1% for character strokes in the closed test, whereas it was 78.9% and 97.7%, respectively, in the open test. We also present a prototype system of the proposed method.

著者

清水 忠昭 木本 雅也 吉村 宏紀 井須 尚紀 菅田 一博

出版者

日本神経回路学会

雑誌

日本神経回路学会誌 (ISSN:1340766X)

巻号頁・発行日

vol.11, no.4, pp.167-175, 2004-12-05 (Released:2011-03-14)

参考文献数

14

---

We showed a new scheme to characterize speech from LSP parameters by 5 layers sandglass type nonlinear neural network (SNN(NL5)). In order to synthesize speech, we take advantage of useful abilities of SNN(NL5) for compressing and restoring the information. We performed learning experiments on LSP parameters of 5 vowels to investigate the ability of SNN. The followings were verified, 1) the distribution of LSP parameters compressed by SNN(NL5) are similar to the distribution of F1-F2 formants plane. 2) Nonlinear output function of neural elements in second and fourth layers of SNN(NL5) work effectively from view point of separating the distribution of vowels. 3) In order to prevent SNN(NL5) from over learning, there exists the optimum numbers of neural elements in second and fourth layers. For 14 orders of LSP parameters, this number was determined to be 20. 4) There is a preferable property on the plane to separate the vowels distinctively when the restoring error of LSP parameters becomes less. 5) SNN(NL5) can restore the LSP parameters with accuracy enough to synthesize speech from the compressed parameters.