On automatic extraction of prosodic information for automatic speech recognition system
On automatic extraction of prosodic information for automatic speech recognition system
复制标题
自动语音识别系统韵律信息的自动提取研究
DOI：
--
复制
发表时间：
1989
期刊：
影响因子：
--
通讯作者：
J. Vaissière
中科院分区：
文献类型：
--
作者：
J. Vaissière
本文关注导致系统中出现错误的三种类型的原因，该系统使用严格与说话人无关的规则，从所测量的法语独立句子的韵律参数（PP）中自动提取语言信息：PP、持续时间和基本频率的错误测量（1 型错误）；不适合相同韵律变化的说话者之间的差异（2 类问题）以及对持续时间的分段影响的某些组合，这些组合不能在严格的自下而上系统中排除（3 类问题）。它表明，现有规则的进一步调整和统计学习都不是完整的解决方案。 1 类错误对于韵律模块来说是外在的，并且很难改进。减少由于类型 2 问题引起的不确定性的一种有效方法是根据说话者的特定习惯部分调整规则集：适应是可行的，因为在韵律模式中存在显着的说话者内部一致性，至少在连续阅读的孤立句子中是这样。 3 类错误 在某些情况下会导致多种解决方案。因此，有必要在一定程度上对说话者之间的变异性进行建模。
This paper is concemed with three types of causes leading to errors in a system using strictly speaker -independent rules for automatic extraction oflinguistic information from measured prosodic parameters (PP) in read isolated sentences, in French: erroneous measurements of PP, duration and fun damental frequency (type-1 errors); differences between speakers who do not fit into the same prosodic moult (type-2 problems) and certain combination of segmental influences on duration, which cannot be factored out in a strictly bottom-up system (type-3 problems). It suggests that neither further tuning ofthe existing rules, nor statisticallearning are complete solutions. Type-1 errors are extrinsic to the prosodic module and can be hardly improved. An effective way of reducing incertainties due to type-2 problems is a partial tuning of the set of rules to the particular habits ofthe speaker: adaptation is feasible because there is a remarkable intra speaker consistency in prosodic patteming, at least in serially read isolated sentences. Type-3 errors Ieads to multiple solutions in certain cases. It is necessary therefore to model to a certain extent the inter-speaker variability.