排序方式: 共有70条查询结果,搜索用时 31 毫秒
1.
Many models of spoken word recognition posit the existence of lexical and sublexical representations, with excitatory and inhibitory mechanisms used to affect the activation levels of such representations. Bottom-up evidence provides excitatory input, and inhibition from phonetically similar representations leads to lexical competition. In such a system, long words should produce stronger lexical activation than short words, for 2 reasons: Long words provide more bottom-up evidence than short words, and short words are subject to greater inhibition due to the existence of more similar words. Four experiments provide evidence for this view. In addition, reaction-time-based partitioning of the data shows that long words generate greater activation that is available both earlier and for a longer time than is the case for short words. As a result, lexical influences on phoneme identification are extremely robust for long words but are quite fragile and condition-dependent for short words. Models of word recognition must consider words of all lengths to capture the true dynamics of lexical activation. (PsycINFO Database Record (c) 2010 APA, all rights reserved) 相似文献
2.
为提高连续语音识别中的音素识别准确率,采用深可信网络提取语音音素后验概率进行音素识别.首先利用受限玻尔兹曼机的学习原理,对深可信网络进行逐层的预训练;然后通过增加一个“软最大化(softmax)”输出层,得到用于音素状态后验概率检测的深层神经网络,并采用后向传播算法进行网络权值的精细调整;最后以后验概率为HMM发射概率,使用Viterbi解码器进行音素识别.针对TIMIT语料库的实验结果表明,该系统的音素识别率优于GMM/HMM,MLP/HMM和TANDEM系统性能. 相似文献
3.
4.
5.
Application of Kernel-Based Feature Space Transformations and Learning Methods to Phoneme Classification 总被引:1,自引:0,他引:1
This paper examines the applicability of some learning techniques to the classification of phonemes. The methods tested were artificial neural nets (ANN), support vector machines (SVM) and Gaussian mixture modeling (GMM). We compare these methods with a traditional hidden Markov phoneme model (HMM), working with the linear prediction-based cepstral coefficient features (LPCC). We also tried to combine the learners with linear/nonlinear and unsupervised/supervised feature space transformation methods such as principal component analysis (PCA), independent component analysis (ICA), linear discriminant analysis (LDA), springy discriminant analysis (SDA) and their nonlinear kernel-based counterparts. We found that the discriminative learners can attain the efficiency of HMM, and that after the transformations they can retain the same performance in spite of the severe dimension reduction. The kernel-based transformations brought only marginal improvements compared to their linear counterparts. 相似文献
6.
汉字改革是社会发展的需要。本文认为音义一体化是汉字发展的方向,也应是汉字改革的方向。而以减少笔画为目的的改革并不能从根本上解决汉字的“三难”问题,因而有必要从整理字素入手,确定出统一规范的音素和义素,并合理调整汉字的结构,进行一次较之历史上的“篆隶之变”更为彻底的综合性改革。 相似文献
7.
基于统计模型及SVM的低速率语音编码QIM隐写检测 总被引:1,自引:0,他引:1
QIM(Quantization Index Modulation,量化索引调制)隐写在标量或矢量量化时嵌入机密信息,可在语音压缩编码过程中进行高隐蔽性的信息隐藏,文中试图对该种隐写进行检测.文中发现该种隐写将导致压缩语音流中的音素分布特性发生改变,提出了音素向量空间模型和音素状态转移模型对音素分布特性进行了量化表示.基于所得量化特征并结合SVM(Support Vector Machine,支持向量机)构建了隐写检测器.针对典型的低速率语音编码标准G.729以及G.723.1的实验表明,文中方法性能远优于现有检测方法,实现了对QIM隐写的快速准确检测. 相似文献
8.
Cross-modal semantic priming and phoneme monitoring experiments investigated processing of word-final nonreleased stop consonants (e.g., kit may be pronounced /kIt/ or /kI/), which are common phonological variants in American English. Both voiced /d/ and voiceless /t/ segments were presented in release and no-release versions. A cross-modal semantic priming task (Experiment 1) showed comparable priming for /d/ and /t/ versions. A second set of stimuli ending in /s/ were presented as intact, missing /s/, or with a mismatching final segment and showed significant but reduced priming for the latter two conditions. Experiment 2 showed that phoneme monitoring reaction time for release and no-release words and onset mismatching stimuli (derived pseudowords) increased as acoustic-phonetic similarity to the intended word decreased. The results suggest that spoken word recognition does not require special mechanisms for processing no-release variants. Rather, the results can be accounted for by means of existing assumptions concerning probabilistic activation that is based on partial activation. (PsycINFO Database Record (c) 2010 APA, all rights reserved) 相似文献
9.
The aims of this study were to investigate the adequacy of electronic voice keys for the purpose of measuring naming latency and to test the assumption that voice key error can be controlled by matching conditions on initial phoneme. Three types of naming latency measurements (hand-coding and 2 types of voice keys) were used to investigate effects of onset complexity (e.g., sat vs. spat) on reading aloud (J. R. Frederiksen & J. F. Kroll, 1976, A. H. Kawamoto & C. T. Kello, 1999). The 3 measurement techniques produced the 3 logically possible results: a significant complexity advantage, a significant complexity disadvantage, and a null effect. Analyses of the performance of each voice key are carried out, and implications for studies of naming latency are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved) 相似文献
10.
Previous research has suggested that the initial portion of a word activates similar sounding words that compete for recognition. Other research has shown that the number of similar sounding words that are activated influences the speed and accuracy of recognition. Words with few neighbors are processed more quickly and accurately than words with many neighbors. The influences of the number of lexical competitors in the initial part of the word were examined in a shadowing and a lexical-decision task. Target words with few neighbors that share the initial phoneme were responded to more quickly than target words with many neighbors that share the initial phoneme. The implications of onset-density effects for models of spoken-word recognition are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved) 相似文献