DEVELOPMENT OF THE SPEECH TRAINING SOFTWARE FOR HEARING IMPAIRED CHILDREN USING THE NETWORK ENVIRONMENT

利用网络环境的听障儿童言语训练软件的开发

• 批准号: 11680229
• 负责人: UMEZAKI Taizo
• 金额: $ 2.05万
• 依托单位: CHUBU UNIVERSITY
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 1999
• 资助国家: 日本
• 起止时间: 1999 至 2001
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-11680229/
• 关键词: SPEECH TRAINING; HEARING IMPAIRED CHILDREN; INTERNET EDUCATION; LIP READING; DEAF; HIDDEN MARKOV MODEL; UTTERANCE EXPRESSION; KOHONEN NEURAL NETWORKS; ゲームランド; ハイブリッド型; 通信教育; Open GL; DPマッチング; 発語訓練; 子音認識; 単語訓練; ニューラルネットワーク

Our (Umezaki Lab) research result of 3 years is as follows.The recognition system of the hybrid type to have integrated three kinds of recognition methods (DP matching, hidden Markov model, artificial neural networks) was developed. With this, the recognition rate of the consonant improved so far more.We developed the utterance training software of the three dimensional version which can be trained in the game sense using the Open GL library. An intellectual training element is incorporated into this software. The title of the software includes the trapshooting and meteorite collision (shooting game), the firework (fantasy game) and so on.We added the following feature to the Internet education system, 1)the data management feature to record a training result, 2)the communication feature that the accumulated training result accumulates to the server through the internet, 3)the report feature to guide the following training based on the training result which, was accumulated by the server.We developed the lip reading training software which can be built with the small-scale database. In this software, we don' t need complicated rule setting. In this software, the image which was produced by the speech synthesis library and the Kohonen-type neural networks is used. The synthesized image which is near the video image is generated by this. The children can train lip reading alone at the house.We are distributing the CD-ROM of the developed software for free. At present, it is already used, in about 60 relation organizations such as the school for the deaf under Aichi Prefecture. Especially, the teachers at the Chikusa and Okazaki schools for the deaf cooperated aggressively.

我们（梅崎实验室）三年的研究成果如下。开发了集成了三种识别方法（DP匹配、隐马尔可夫模型、人工神经网络）的混合型识别系统。至此，辅音的识别率有了较大的提高。我们开发了三维版本的语音训练软件，可以利用Open GL库进行游戏意义上的训练。该软件中融入了智力培训元素。软件名称包括陷阱射击、陨石碰撞（射击游戏）、烟花（奇幻游戏）等。我们在互联网教育系统中增加了以下功能，1）记录训练结果的数据管理功能，2）将累积的训练结果通过互联网积累到服务器的通讯功能，3）报告功能，根据服务器累积的训练结果指导后续的训练。我们开发了唇读训练软件，可以用以下软件构建 小型数据库。在这个软件中，我们不需要复杂的规则设置。在该软件中，使用了由语音合成库和 Kohonen 型神经网络生成的图像。由此生成接近视频图像的合成图像。孩子们可以在家里独自训练唇语。我们免费分发开发软件的光盘。目前，已在爱知县聋人学校等约60个相关机构中使用。尤其是千草聋哑学校和冈崎聋哑学校的老师们积极配合。

项目成果

木村成宏, 梅崎太造, 佐藤省三: "母音同期に基づく発話表情画像の自動生成"電気関係学会,東海支部連合大会,講論集. 505. 253 (2001)

Shigehiro Kimura、Taizo Umezaki、Shozo Sato：“基于元音同步的语音表达图像的自动生成”，电气工程学会，东海分会会议，论文集 505. 253 (2001)。

猪又信義, 梅崎太造, 田中泰史: "3D版発話訓練ソフトウェアの開発"電気関係学会,東海支部連合大会,講論集. 506. 253 (2001)

Nobuyoshi Inomata、Taizo Umezaki、Yasushi Tanaka：“3D版语音训练软件的开发”，电气工程学会，东海分会联盟会议，会议记录506。253（2001）。

梅崎太造,間瀬寛子,江川敦史: "ゲーム性を取り入れた聴覚障害児用発話訓練ソフトウェアの開発"日本福祉工学会誌. Vol.2. 31-36 (2000)

Taizo Umezaki、Hiroko Mase、Atsushi Ekawa：“开发具有游戏功能的听力障碍儿童语音训练软件”日本福利工程学会杂志第 2 卷 31-36（2000 年）。

近田智洋, 高橋久, 梅崎太造: "聴覚障害児用発話訓練システム"電子情報通信学会,信学技報. WIT00-4. 17-22 (2000)

Tomohiro Chikada、Hisashi Takahashi、Taizo Umezaki：“听力障碍儿童的言语训练系统”，电子、信息和通信工程师协会，IEICE 技术报告 17-22 (2000)。

TAIZO UMEZAKI, HIROKO MASE, ATSUSHI EGAWA: "DEVELOPMENT OF SPEECH TRAINING SOFTWARE FOR HEARING IMPAIRED CHILDREN WHICH INTRODUCED A GAME"THE JOURNAL OF THE JAPAN SOCIETY FOR WELFARE ENGINEERING. VOL.2. 31-36 (2000)

TAIZO UMEZAKI、HIROKO MASE、ATSUSHI EGAWA：“为听力障碍儿童开发引入游戏的言语训练软件”日本福祉工程学会杂志。