会员体验
专利管家(专利管理)
工作空间(专利管理)
风险监控(情报监控)
数据分析(专利分析)
侵权分析(诉讼无效)
联系我们
交流群
官方交流:
QQ群: 891211   
微信请扫码    >>>
现在联系顾问~
热词
    • 1. 发明公开
    • 스위칭 상태 공간 모델들을 갖는 변분 추론을 사용하는음성 인식 방법
    • 使用具有切换状态空间模型的变异语音的语音识别方法
    • KR1020040088368A
    • 2004-10-16
    • KR1020040022168
    • 2004-03-31
    • 마이크로소프트 코포레이션
    • 아티아스하가이리레오진규뎅리
    • G10L15/14
    • G10L15/14
    • PURPOSE: A speech recognition method using variational inference having switching state spatial models is provided to allow boundaries of phonetic units to be trained with parameters of HDM(Hidden Dynamic Model) other than boundary parameters and overcome processing incapability related to the training. CONSTITUTION: A speech recognition method includes a step of defining a hidden generation-related parameter and a switching state spatial model, and a step of approximating posterior probability providing possibility of the sequence of phonetic units and the sequence of the hidden generation-related parameters based on the sequence of the input values. The posterior probability approximating step optimizes the boundaries of the phonetic units.
    • 目的:提供一种使用具有开关状态空间模型的变分推理的语音识别方法,以允许使用边界参数以外的HDM(隐藏动态模型)参数训练语音单元的边界,并克服与训练相关的处理不适用性。 构成:语音识别方法包括定义隐藏生成相关参数和切换状态空间模型的步骤,以及近似提供语音单元序列的可能性的后验概率和隐藏生成相关参数的序列的步骤 关于输入值的顺序。 后验概率近似步骤优化了语音单元的边界。