發(fā)簡信
IP屬地：江蘇
-
Voice ConversionIntroduction VC aims to convert the non-linguistic information of the sp...
-
ASR SystemsIntroduction The ASR system can be categoried as three classes by its ou...
-
Language Model for ASRBackground Automatic Speech Recognition (ASR) uses both acoustic model (...
-
語言模型融合 Language Model Fusion
Introduction In the previous articals, we have learnt the CTC loss makes...
-
Keyword Spotting關(guān)鍵詞偵聽Introduction Keyword Spotting (KWS) aims at detecting predefined key-wor...
-
注意力機制的增強Enhancement of Attention MechanismMulti-headed Attention 一個attention head可能權(quán)重大部分在某處坟岔，不能提取豐富的信息，需要多個進行融合钓觉。 Fu...
-
Query-Key-Value Perspective on Attention Mechanism 怎么用“查詢-鍵-值”理解注意力機制注意力機制 RNN編碼-解碼模型 論文[1]中衫贬，從RNN編碼-解碼模型演進出注意力機制朝氓。RNN編碼-解碼模型中缸逃，編碼器輸入序列少辣，是編碼器RNN在...
-
CTC序列模型背景 手寫體識別、語音識別中笤成，輸入數(shù)據(jù)和輸出的識別結(jié)果長度不一致评架、而且可變。直接用神經(jīng)網(wǎng)絡(luò)訓(xùn)練需要預(yù)分割炕泳、調(diào)整纵诞，得到對應(yīng)關(guān)系，這很難做到培遵。CTC...
-
Faster R-CNN網(wǎng)絡(luò)架構(gòu) 可以分為3個部分 Head Region Proposal Network(RPN) Classification Network R...