Overview
Who this competition fits
Builders comfortable with Chinese algorithm-competition pages hosted by iFlytek.
Read the original official blurb
目前的唇语识别技术依赖于语言模型对说话人发音连续的话语进行识别。然而，当面对发音不连续的语句时，这种方法无法有效地识别出说话人所说的内容。为了应对这种情况，我们采用了关键词识别的方法，不依赖于语言模型，获取了每个拼音在视频每一帧中存在的概率矩阵以及每个拼音在视频中是否存在的概率矩阵。基于这两个概率矩阵，旨在进行一次语句的识别，从而预测出完整且连续的一句… Official track: 语音技术. Sponsor: 合肥工业大学. Teams entered: 140.
Preparation
From registration to a first submission
- 01
Can read Chinese competition pages
- 02
Comfortable submitting a baseline
Before you commit: Task details, eligibility, and scoring rules live mainly on the official iFlytek page and change each season.
Source
How this page was assembled
Competition information is structured from the official page. Scores are platform estimates for decision support; official rules take precedence.
- Official competition page
- iFlytek Challenge
- Last checked
- May 14, 2024