Bayesian background models for keyword spotting in handwritten documents
作者:
Highlights:
• We propose Bayesian background models for keyword spotting in handwritten documents.
• We cover a detailed illustration of two of the proposed Bayesian background models.
• The Bayesian formulation adds uncertainty to handle variation in writing styles.
• The weights learned on the individual samples provide better rejection criteria.
• A line level approach that avoids any error is introduced by the word segmentation.
摘要
Highlights•We propose Bayesian background models for keyword spotting in handwritten documents.•We cover a detailed illustration of two of the proposed Bayesian background models.•The Bayesian formulation adds uncertainty to handle variation in writing styles.•The weights learned on the individual samples provide better rejection criteria.•A line level approach that avoids any error is introduced by the word segmentation.
论文关键词:Handwriting recognition,Keyword spotting,Bayesian generalized linear models,Bayesian generalized kernel models
论文评审过程:Received 3 February 2015, Revised 24 May 2016, Accepted 29 June 2016, Available online 15 July 2016, Version of Record 9 November 2016.
论文官网地址:https://doi.org/10.1016/j.patcog.2016.06.030