Infant behavioral state and speech sound discrimination

In order to investigate the interaction between infant state and speech sound discrimination, records of state were obtained from infants participating in 3 speech sound discrimination experiments which employed the high amplitude sucking procedure. A total of 94 individual records were rated for state by 2 independent scorers. When treated as an independent variable, state interacted with responsivity to sound change as measured by sucking rate increases. Infants in the more alert states demonstrated significantly higher rates of sucking following sound change. Treating state as a dependent variable revealed that state was not significantly affected by a change in stimulus conditions.     

语音识别系统软件设计

提取语音信号的MFCC特征参数,用矢量量化（VQ）的LBG算法来建立匹配模板.在MATLAB软件平台上,基于GUI界面实现说话人识别系统软件设计,并通过实验验证其有效性.     

语音情感识别分类算法研究综述

语音情感识别本质上是对语音情感的特征参数进行分类和模式识别,其核心问题在于寻找一个更为合适的语音情感识别算法.综述了语音情感识别算法的分类概况、常用的分类算法、研究中的困难与需要进一步研究的问题.最后展望了可能提高精度的先进识别模型.     

基于支持向量机的语音情感识别

针对语音情感识别特征识别问题，本利用支持向量机进行了研究．分析表明语音信号的情感特征参数在输入空间中不完全是一个线性分类的问题，使用非线性的核函数对输入空间进行映射可以有效地提高识别效率．与已有的多模式语音情感识别方式相比，利用高斯(径向基)核函数的支持向量机的识别效果优于其他已有的方法．     

基于HMM状态结构调整的非特定人语音识别

利用HMM模型状态间的混淆度, 提出了一种新的状态结构调整算法, 使不同的状态可以共享相同的高斯混合函数, 并在EM算法的框架下推导出对状态结构调整后的增加参数, 即状态间权值的重估公式. 并对非特定人进行大词汇量汉语连续语音识别实验, 实验结果表明状态结构调整后的系统不仅优于基线系统, 还获得了比传统的参数增加方法更高的识别率, 由此证明了状态结构调整方法的有效性.     

How do children read multisyllabic words? Some preliminary observations

Multisyllabic words have been neglected in determining the relationship between spelling and sound in reading development. In a preliminary exploration of this topic, sensitivity to the phonological and orthographic composition of multisyllabic words and nonwords is examined amongst a group of English‐speaking 11‐year‐olds. The nature of the English language suggests that the syllable structure and stress pattern of words may influence the acquisition of higher‐order reading skills. A phonological awareness task confirms that syllable boundaries are ambiguous in certain English words. Furthermore, accuracy at reading multisyllabic words and nonwords appears sensitive to this ambiguity as a small advantage emerges for stimuli with more stable syllable structures. Nonword but not word reading is affected by syllable length, and nonwords are assigned stress patterns which appear to be related to the lexical syllables that were used to construct these items. These findings are related to current connectionist models of word recognition.     

机器人语音识别实验设计与实现

基于Matlab设计了以动态时间规整(dynamic time warping,DTW)算法和Mel频率倒谱系数(Mel frequency cepstrum coefficients,MFCC)参数提取算法为基础的孤立词自动语音识别系统;然后结合机器人语音识别的需求,基于凌阳SPCE061A设计了语音识别应用系统。结合上述两部分工作,设计、实现了机器人语音识别系统演示实验和机器人语音识别半开放实验,实现效果达到预期实验设计目标。     

航管仿真系统中语音技术的应用

分析了在航管雷达模拟系统增加语音识别和合成功能的实现过程。根据航管雷达模拟系统的特点,采用中英文皆可识别且适合连续识别的语音引擎,以及符合人类语言特点的TTS语音库,与系统内部通讯机制相结合,实现语音识别和合成的过程。该系统运行稳定,在识别率和数据传输速度等方面效果比较好。     

汉语语音识别中区分性声调模型及最优集成方法

提出了2种解决汉语语音识别中声调问题的方法:利用区分性方法对基于隐马尔可夫模型(HMM)的声调模型进行训练;提出将区分性训练的声调模型加入大词汇量连续语音识别系统的最优方法,该方法根据最小音子错误的训练准则以及利用扩展Baum-Welch算法区分性训练与模型相关的概率权重,对声学模型以及声调模型概率进行加权.实验结果表明区分性训练的声调模型能够显著地提高连续语音声调识别率以及大词汇量语音识别系统的识别率,同时区分性的模型权重训练能够在区分性声调模型加入连续语音识别系统之后进一步提高系统的识别性能.     

Application of formant instantaneous characteristics to speech recognition and speaker identification

This paper proposes a new phase feature derived from the formant instantaneous characteristics for speech recognition (SR) and speaker identification (SI) systems. Using Hilbert transform (HT), the formant characteristics can be represented by instantaneous frequency (IF) and instantaneous bandwidth, namely formant instantaneous characteristics (FIC). In order to explore the importance of FIC both in SR and SI, this paper proposes different features from FIC used for SR and SI systems. When combing these new features with conventional parameters, higher identification rate can be achieved than that of using Mel-frequency cepstral coefficients (MFCC) parameters only. The experiment results show that the new features are effective characteristic parameters and can be treated as the compensation of conventional parameters for SR and SI.     

Pavlovian conditioning in multiple contexts: Competition between contexts for comparator status

Water-deprived rats were used to investigate the effects of training a CS in more than one context on conditioned lick suppression. In each experiment, partial reinforcement of the CS was intermingled with unsignaled presentations of the US. In Experiment 1, subjects were either trained in one context alone, trained consecutively in two contexts (such that all training in one context occurred prior to any training in the second context), or trained alternately in two contexts. Following training, the first context, the second context, or neither context was extinguished. Testing of the CS occurred in a third (neutral) context. To the extent that either training context became established as a comparator stimulus for the CS, the comparator hypothesis (Miller & Matzel, 1988) predicts an increase in excitatory responding to the CS following extinction of that context. Subjects trained in a single context exhibited appreciable fear of the CS only when the CS’s training context had been extinguished. Additionally, subjects trained consecutively in the two contexts showed increased fear of the CS following extinction of the second, but not the first training context (i.e., a recency effect). Subjects trained alternately in the two contexts showed no increased fear of the CS as a result of either context alone being extinguished. In Experiment 2, subjects trained alternately in two contexts showed increased fear of the CS only when both training contexts were extinguished, suggesting that both training contexts had become comparator stimuli. These data indicate that multiple training contexts can either compete or act synergistic-ally in modulating responding to a Pavlovian trained CS as a function of the order of training in the different contexts.