针对现有语音关键词检测方法定位精度低的问题,提出了一种基于多尺度距离矩阵的语音关键词检测与细粒度定位方法(spoken term detection and fine-grained localization method based on multi-scale distance matrices,MF-STD)。该方...针对现有语音关键词检测方法定位精度低的问题,提出了一种基于多尺度距离矩阵的语音关键词检测与细粒度定位方法(spoken term detection and fine-grained localization method based on multi-scale distance matrices,MF-STD)。该方法首先利用残差卷积网络提取特征并构建距离矩阵以建模输入之间的相关性;其次通过多尺度分割和解耦头学习不同尺度下的定位信息;最后根据多尺度加权定位损失、置信度损失和分类损失优化模型,实现对关键词存在性和时域边界的细粒度预测。在LibriSpeech数据集上的实验结果表明,MF-STD在集内词的检测中,精准率和交并比分别达到97.1%和88.6%;在集外词的检测中,精准率和交并比分别达到96.7%和88.2%。与现有的语音关键词检测与定位方法相比,MF-STD的检测准确率和定位精度显著提升,充分证明该方法的先进性,也证明了多尺度特征建模与细粒度定位约束在语音关键词检测任务中的有效性。展开更多
An important component of a spoken term detection （STD） system involves estimating confidence measures of hypothesised detections.A potential problem of the widely used lattice-based confidence estimation,however,is...An important component of a spoken term detection （STD） system involves estimating confidence measures of hypothesised detections.A potential problem of the widely used lattice-based confidence estimation,however,is that the confidence scores are treated uniformly for all search terms,regardless of how much they may differ in terms of phonetic or linguistic properties.This problem is particularly evident for out-of-vocabulary （OOV） terms which tend to exhibit high intra-term diversity.To address the impact of term diversity on confidence measures,we propose in this work a term-dependent normalisation technique which compensates for term diversity in confidence estimation.We first derive an evaluation-metric-oriented normalisation that optimises the evaluation metric by compensating for the diverse occurrence rates among terms,and then propose a linear bias compensation and a discriminative compensation to deal with the bias problem that is inherent in lattice-based confidence measurement and from which the Term Specific Threshold （TST） approach suffers.We tested the proposed technique on speech data from the multi-party meeting domain with two state-ofthe-art STD systems based on phonemes and words respectively.The experimental results demonstrate that the confidence normalisation approach leads to a significant performance improvement in STD,particularly for OOV terms with phonemebased systems.展开更多
针对关键词中的集外词检索任务,提出采用音素、音节、词片三种子词单元进行多流信息的联合检索算法,其中对基于音素的语音检索(Spoken term detection,STD)系统使用基于n元语言模型-加权有限状态机的完全匹配检索降低漏警,对基于音节、...针对关键词中的集外词检索任务,提出采用音素、音节、词片三种子词单元进行多流信息的联合检索算法,其中对基于音素的语音检索(Spoken term detection,STD)系统使用基于n元语言模型-加权有限状态机的完全匹配检索降低漏警,对基于音节、词片的STD系统使用模糊匹配检索降低虚警,最后采用线性逻辑回归(Linear logistic regression,LLR)的算法将三个子系统的结果进行融合。在NIST STD 2006语音检索评测的英语电话会话语音测试集上的实验结果表明,相对于最好的单流系统,多流信息融合获得了12%的实际词项权重值(Actual term weighted value,ATWV)相对提升。展开更多
文摘针对现有语音关键词检测方法定位精度低的问题,提出了一种基于多尺度距离矩阵的语音关键词检测与细粒度定位方法(spoken term detection and fine-grained localization method based on multi-scale distance matrices,MF-STD)。该方法首先利用残差卷积网络提取特征并构建距离矩阵以建模输入之间的相关性;其次通过多尺度分割和解耦头学习不同尺度下的定位信息;最后根据多尺度加权定位损失、置信度损失和分类损失优化模型,实现对关键词存在性和时域边界的细粒度预测。在LibriSpeech数据集上的实验结果表明,MF-STD在集内词的检测中,精准率和交并比分别达到97.1%和88.6%;在集外词的检测中,精准率和交并比分别达到96.7%和88.2%。与现有的语音关键词检测与定位方法相比,MF-STD的检测准确率和定位精度显著提升,充分证明该方法的先进性,也证明了多尺度特征建模与细粒度定位约束在语音关键词检测任务中的有效性。
文摘An important component of a spoken term detection （STD） system involves estimating confidence measures of hypothesised detections.A potential problem of the widely used lattice-based confidence estimation,however,is that the confidence scores are treated uniformly for all search terms,regardless of how much they may differ in terms of phonetic or linguistic properties.This problem is particularly evident for out-of-vocabulary （OOV） terms which tend to exhibit high intra-term diversity.To address the impact of term diversity on confidence measures,we propose in this work a term-dependent normalisation technique which compensates for term diversity in confidence estimation.We first derive an evaluation-metric-oriented normalisation that optimises the evaluation metric by compensating for the diverse occurrence rates among terms,and then propose a linear bias compensation and a discriminative compensation to deal with the bias problem that is inherent in lattice-based confidence measurement and from which the Term Specific Threshold （TST） approach suffers.We tested the proposed technique on speech data from the multi-party meeting domain with two state-ofthe-art STD systems based on phonemes and words respectively.The experimental results demonstrate that the confidence normalisation approach leads to a significant performance improvement in STD,particularly for OOV terms with phonemebased systems.
文摘针对关键词中的集外词检索任务,提出采用音素、音节、词片三种子词单元进行多流信息的联合检索算法,其中对基于音素的语音检索(Spoken term detection,STD)系统使用基于n元语言模型-加权有限状态机的完全匹配检索降低漏警,对基于音节、词片的STD系统使用模糊匹配检索降低虚警,最后采用线性逻辑回归(Linear logistic regression,LLR)的算法将三个子系统的结果进行融合。在NIST STD 2006语音检索评测的英语电话会话语音测试集上的实验结果表明,相对于最好的单流系统,多流信息融合获得了12%的实际词项权重值(Actual term weighted value,ATWV)相对提升。
