k2-fsa / icefall

https://k2-fsa.github.io/icefall/
Apache License 2.0
792 stars 267 forks source link

kaldi经典的强制对齐算法怎么在k2实现呢 #1574

Closed ssf008 closed 1 month ago

ssf008 commented 1 month ago

https://github.com/tbright17/kaldi-dnn-ali-gop

对齐每个字 每个音素准确度都不错

不像ctc对齐不准,ctc的后验都是尖峰,对齐不出静音

还是有很强的需求 在帧级别的语音识别和语音合成任务中应用广泛,同时也是字幕自动打轴、口语评测等任务中的核心算法 也可以在自定义唤醒词 强制对齐每个字,音素等

kaldi第一代在强制对齐和发音评测的优秀,k2怎么没有了

JinZr commented 1 month ago

目前还没有计划实现 1st gen Kaldi 里的 force align 算法

Best Regards Jin

On Tue, 2 Apr 2024 at 11:38 ssf008 @.***> wrote:

https://github.com/tbright17/kaldi-dnn-ali-gop

对齐每个字 每个音素准确度都不错

还是有很强的需求

比如自定义唤醒词 强制对齐每个字,音素等

发音评测,对每个字,单词打分

— Reply to this email directly, view it on GitHub https://github.com/k2-fsa/icefall/issues/1574, or unsubscribe https://github.com/notifications/unsubscribe-auth/AOON42D2IIEAI2KFGHKSVVDY3IR5FAVCNFSM6AAAAABFSS7ZE6VHI2DSMVQWIX3LMV43ASLTON2WKOZSGIYTSNBZGE3TQNY . You are receiving this because you are subscribed to this thread.Message ID: @.***>