End-To-End Speech to Braille Translation in Japanese
End-To-End Speech to Braille Translation in Japanese
复制标题
日语盲文翻译的端到端语音翻译
DOI：
10.1109/icce53296.2022.9730468
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Norihide Kitaoka
中科院分区：
文献类型：
--
作者：
Akio Kobayashi ; Junji Onishi ; Hiromitsu Nishizaki ; Norihide Kitaoka
本研究提出一个端到端的盲文翻译方法，从日本的讲话，为盲人。在日本，从口语中自动翻译盲文有望改善盲人获取信息的机会。日语盲文对于自动语音识别（ASR）具有高亲和力，因为其主要包括反映日语语音特征的平假名字符（假名）。因此，我们尝试使用神经网络以端到端（E2E）的方式直接从语音翻译日语盲文。我们还比较了我们提出的E2E方法与现有的方法，结合了ASR和自动盲文翻译。
This study addresses an end-to-end braille translation approach from Japanese speech for the deaf-blind. In Japan, automatic Braille translation from spoken language is expected to improve information accessibility for deaf-blind people. Japanese Braille has a high affinity for automatic speech recognition (ASR) because it primarily comprises hiragana characters (kana) reflecting Japanese phonetic features. Therefore, we attempted to use neural networks to translate Japanese Braille directly from speech in an end-to-end (E2E) manner. We also compared our proposed E2E approach with an existing method that combines the ASR and automatic Braille translation.