倒谱参数稀疏分解下的汉语音谎言检测

为了提高汉语语音的谎言检测准确率,提出了一种对信号倒谱参数进行稀疏分解的方法。首先,采用小波包滤波器组对语音信号进行多频带划分,求得子频带对数能量并进行离散余弦变换以提取小波包频带倒谱系数,结合梅尔频率谱系数得到倒谱参数;其次,依据K-奇异值分解方法分别利用说谎和非说谎两种状态下的语音倒谱参数集训练得到过完备混合字典,在此字典上根据正交匹配追踪算法对参数集进行稀疏编码提取稀疏特征;最终进行多种分类模型下的识别实验·实验结果表明,稀疏分解方法相比传统参数降维方法具有更好的优化性能,本文推荐的稀疏谱特征最佳识别率达到78.34%,优于其他特征参数,显著提高了谎言检测识别准确率。      

基于稀疏表示的核素能谱特征提取及核素识别

提出了一种基于稀疏表示的核素能谱特征提取方法,其实质是将核素能谱在区分性最好的稀疏原子上进行投影。利用稀疏分解方法对核素能谱进行稀疏分解,提取分解系数向量作为表征核素的特征向量,通过模式识别分类方法建立分类模型实现核素识别。与传统稀疏分解方法的区别在于:在能谱稀疏分解过程中按照稀疏字典中的原子排列顺序顺次进行分解;其次,分解目的在于特征提取,即最终提取到的特征对不同核素具有可区分性,并不要求核素能谱的重构精度。在241Am, 133Ba, 60Co, 137Cs, 131I和152Eu共6种核素1200个能谱数据上进行了核素识别实验,7种不同分类算法的平均识别率达到91.71%,实验结果的统计分析表明,本文提出的特征提取方法识别准确率显著地高于两种传统核素能谱特征提取方法准确率。     

Features of underwater echo extraction based on signal sparse decomposition

In order to better realize sound echo recognition of underwater materials with heavily uneven surface,a features ion method based on the theory of signal sparse decomposition has been proposed.Instead of the common time frequency dictionary,sets of training echo samples are used directly as dictionary to realize echo sparse decomposition under L₁ optimization and a kind of energy features of the echo.Experiments on three kinds of bottom materials including the Cobalt Crust show that the Fisher distribution with this method is superior to that of edge features and of Singular Value Decomposition （SVD） features in wavelet domain.It means no doubt that much better classification result of underwater bottom materials can be obtained with the proposed energy features than the other two.It is concluded that echo samples used as a dictionary is feasible and the class information of echo introduced by this dictionary can help to obtain better echo features.     

Statistical modeling of speech Poincaré sections in combination of frequency analysis to improve speech recognition performance

This paper introduces a combinational feature extraction approach to improve speech recognition systems. The main idea is to simultaneously benefit from some features obtained from Poincare? section applied to speech reconstructed phase space (RPS) and typical Mel frequency cepstral coefficients (MFCCs) which have a proved role in speech recognition field. With an appropriate dimension, the reconstructed phase space of speech signal is assured to be topologically equivalent to the dynamics of the speech production system, and could therefore include information that may be absent in linear analysis approaches. Moreover, complicated systems such as speech production system can present cyclic and oscillatory patterns and Poincare? sections could be used as an effective tool in analysis of such trajectories. In this research, a statistical modeling approach based on Gaussian mixture models (GMMs) is applied to Poincare? sections of speech RPS. A final pruned feature set is obtained by applying an efficient feature selection approach to the combination of the parameters of the GMM model and MFCC-based features. A hidden Markov model-based speech recognition system and TIMIT speech database are used to evaluate the performance of the proposed feature set by conducting isolated and continuous speech recognition experiments. By the proposed feature set, 5.7% absolute isolated phoneme recognition improvement is obtained against only MFCC-based features.     

A structural health monitoring strategy using cepstral features

A statistical pattern recognition based damage detection algorithm is proposed. The algorithm is developed according to the training and testing scheme, typical of pattern recognition applications. The original contribution of the work is given by the use of an adaptation of Mel-Frequency Cepstral Coefficients as damage sensitive features, as their compactness and de-correlation characteristics make them particularly suited for statistical pattern recognition applications. At the same time, the ease of extraction, which requires minimal user expertise, represents an important advantage over other more popular features, and makes the cepstral features particularly convenient for implementation into automatic structural health monitoring routines. The damage detection algorithm employs the squared Mahalanobis distance to solve the Structural Health Monitoring assignment. The method is validated by using both simulated and experimental data, and the performance of said features is compared to that of Auto-Regressive (AR) coefficients, which have been largely used to solve the task of structural damage detection. The experimental data were measured on a steel frame, which behave nonlinearly in its damaged configuration, at the Los Alamos National Laboratory. Results demonstrate that the proposed approach may be conveniently used in real-life applications, since cepstral features outperform AR coefficients when dealing with experimental data modeled to mimic the operational and environmental variability.     

稀疏低秩噪声模型下无监督实时单通道语音增强算法

针对现有基于字典学习的增强算法需要先验信息、不易实时处理的问题,提出一种便于实时处理的无监督的单通道语音增强算法。首先,该算法将无监督条件下背景噪声的建模问题转化为带噪语音幅度谱的稀疏低秩噪声分解;然后,采用增量非负子空间方法对背景噪声进行在线字典学习,获得能够体现背景噪声时变特性的自适应噪声字典;最后,利用所得的噪声字典,采用易于实时处理的逐帧迭代方式,对带噪语音进行处理。实验结果表明:相较于多带谱减法和基于低秩稀疏矩阵分解的增强算法,所提算法在噪声抑制方面的性能尤为显著,在多项性能评价指标上,均表现出更好的结果。      

卷积噪声环境下语音信号鲁棒特征提取

提出了一种基于独立分量分析(ICA)的语音信号鲁棒特征提取算法,用以解决在卷积噪声环境下语音信号的训练与识别特征不匹配的问题。该算法通过短时傅里叶变换将带噪语音信号从时域转换到频域后,采用复值ICA方法从带噪语音的短时谱中分离出语音信号的短时谱,然后根据所得到的语音信号短时谱计算美尔倒谱系数(MFCC)及其一阶差分作为特征参数。在仿真与真实环境下汉语数字语音识别实验中,所提算法相比较传统的MFCC其识别正确率分别提升了34.8%和32.6%。实验结果表明基于ICA方法的语音特征在卷积噪声环境下具有良好的鲁棒性。      

基于稀疏分解的大功率半导体激光器1/f噪声参数估计的新方法

本文结合1/f噪声信号功率谱随频率成反比变化的关系, 以及稀疏分解可以根据信号灵活构造原子库的特点, 提出一种基于稀疏分解估计大功率半导体激光器1/f噪声的新方法, 构造了具备1/f噪声特点的过完备库. 在该过完备库中通过Matching Pursuit(MP)算法完成了白噪声与1/f噪声混叠信号的稀疏分解. 实验结果显示:该方法估计出淹没在白噪声环境中1/f噪声的γ 参数, 与频谱分析仪的测量结果有较好的一致性, 通过对比不同的过完备库证明了所构造的过完备库的优越性.     

Classification of stop place in consonant-vowel contexts using feature extrapolation of acoustic-phonetic features in telephone speech

Knowledge-based speech recognition systems extract acoustic cues from the signal to identify speech characteristics. For channel-deteriorated telephone speech, acoustic cues, especially those for stop consonant place, are expected to be degraded or absent. To investigate the use of knowledge-based methods in degraded environments, feature extrapolation of acoustic-phonetic features based on Gaussian mixture models is examined. This process is applied to a stop place detection module that uses burst release and vowel onset cues for consonant-vowel tokens of English. Results show that classification performance is enhanced in telephone channel-degraded speech, with extrapolated acoustic-phonetic features reaching or exceeding performance using estimated Mel-frequency cepstral coefficients (MFCCs). Results also show acoustic-phonetic features may be combined with MFCCs for best performance, suggesting these features provide information complementary to MFCCs.     

Second generation wavelet transform-based pitch period estimation and voiced/unvoiced decision for speech signals

Pitch detection is an important part of speech recognition and speech processing. In this paper, a pitch detection algorithm based on second generation wavelet transform was developed. The proposed algorithm reduces the computational load of those algorithms that were based on classical wavelet transform. The proposed pitch detection algorithm was tested for both real speech and synthetic speech signal. Some experiments were carried out under noisy environment condition to evaluate the accuracy and robustness of the proposed algorithm. Results showed that the proposed algorithm was robust to noise and provided accurate estimates of the pitch period for both low-pitched and high-pitched speakers. Moreover, different wavelet filters that were obtained using second generation wavelet transform were considered to see the effects of them on the proposed algorithm. It was noticed that Haar filter showed good performance as compared to the other wavelet filters.     

基于信号稀疏分解的水下回波分类

针对表面起伏不平的水下底质回波分类效果差的问题,提出一种新颖的基于信号稀疏分解理论的水下底质回波特征提取方法。本方法并不使用通用时频字典,而是针对回波分类这一中心任务直接采用回波训练样本集作为字典,将水下回波信号在该字典上进行稀疏分解,然后提取出回波信号的类别能量特征。对水下钴结壳等三类底质回波分类实验表明,基于信号稀疏分解的类别能量特征的fisher分布明显优于小波域模极大值边缘特征和奇异值特征,从而显著提高了水下回波的分类效果。研究结论:在回波特征提取阶段,采用回波样本作为信号表达字典是可行的,同时由回波样本字典引入的回波类别信息将有助于获取更优的回波特征。      

基于变分模态分解的语音情感识别方法*

针对传统语音情感特征参数在进行情感分类时性能不佳的问题,该文提出了一种基于变分模态分解的语音情感识别方法。情感语音信号首先由变分模态分解提取固有模态函数,然后对所选主导固有模态函数进行重新聚合,再提取梅尔倒谱系数和各固有模态函数的希尔伯特边际谱。为了验证该文提出的特征性能,选用两种语音数据库(EMODB、RAVDESS)进行实验,按该文方法提取特征后使用极限学习机进行语音情感分类识别。实验结果表明:相比基于经验模态分解和集合经验模态分解的语音情感特征,该文提出的特征有更好的识别性能,验证了该方法的实用性。     

结合幅度谱和功率谱字典的语音增强方法

从双路字典学习、噪声功率谱估计、语音幅度谱重构角度提出了一种改进的谱特征稀疏表示语音增强方法。在字典学习阶段,融合功率谱与幅度谱特征,采用区分性字典降低语音字典和噪声字典的相干性;在语音增强阶段,提出一种噪声功率谱估计方法对非平稳噪声进行跟踪估计;考虑到幅度谱和功率谱特征对不同噪声的适应程度不同,设计了语音重构权值表。对分别由幅度谱和功率谱恢复而来的两路信号进行自适应加权重构,结合相位补偿函数得到增强后的语音信号。实验结果表明,该方法在平稳、非平稳噪声环境下相比于单一谱特征的语音增强方法平均提高31.6%,改善了语音增强方法的性能。      

Spectro-temporal modulation energy based mask for robust speaker identification

Spectro-temporal modulations of speech encode speech structures and speaker characteristics. An algorithm which distinguishes speech from non-speech based on spectro-temporal modulation energies is proposed and evaluated in robust text-independent closed-set speaker identification simulations using the TIMIT and GRID corpora. Simulation results show the proposed method produces much higher speaker identification rates in all signal-to-noise ratio (SNR) conditions than the baseline system using mel-frequency cepstral coefficients. In addition, the proposed method also outperforms the system, which uses auditory-based nonnegative tensor cepstral coefficients [Q. Wu and L. Zhang, "Auditory sparse representation for robust speaker recognition based on tensor structure," EURASIP J. Audio, Speech, Music Process. 2008, 578612 (2008)], in low SNR (≤ 10 dB) conditions.     

A united optimum images fusion based on analysis of color distortion

In remote sensing community, IHS (intensity, hue, and saturation) transform is one of the most commonly used fusion algorithm. A study on IHS fusion indicates that the color distortion cannot be avoided.Meanwhile, wavelet decomposition has a property of frequency division in transform domain. And the statistical property of wavelet coefficient reflects those significant features. So, a united optimal fusion method, which using the statistical property of wavelet decomposition and IHS transform on pixel and feature levels, is proposed. That is, the high frequency of intensity component I is fused on feature level with multi-resolution wavelet in IHS space, and the low frequency of intensity component I is fused on pixel level with optimal weight coefficients. Spectral information and spatial resolution are two performance indexes of optimal weight coefficients. Experiment results show that it is a practical and effective method.     

匹配追踪说话人自适应方法

针对现有子空间自适应方法无法确定最佳说话人子空间的问题,提出一种基于匹配追踪的说话人自适应方法。将说话人自适应视为一种高维信号的稀疏分解问题,利用本征音和参考说话人超矢量的各自优势联合构造说话人字典;依据匹配追踪原理,通过迭代优化,以后验方式确定最佳说话人子空间维数及其基矢量。引入冗余基矢量检测与去除机制以保证算法的稳定性,并通过快速递推算法得到新说话人坐标。基于汉语连续语音识别的有监督说话人自适应实验结果表明,与本征音及参考说话人加权方法相比,平均有调音节正识率相对提高了1.9%。      

联合稀疏表示在水下目标分类中的应用

水下目标分类识别的性能受所选特征的限制,多特征往往可以获得更加稳定的结果,针对这一问题,提出了一种基于联合稀疏表示模型的水下目标分类识别方法。首先对水下目标回波信号提取3种具有信息互补性与关联性的特征:中心矩特征、小波包能量谱特征、梅尔频率倒谱系数特征,然后应用加速近端梯度法对联合稀疏表示模型进行优化,求解得到最优联合稀疏系数,最后根据最小误差准则确定目标类别。在消声水池开展模拟实验,对6类目标进行分类识别,结果表明:与传统算法相比,提出的算法具有更高识别准确率,并且其执行效率较传统算法有很大提升。      

基于可见光的多波段偏振图像融合新算法

采用了一种新的基于小波变换的偏振图像融合算法.首先,将两个波段中的每一波段三幅偏振图像利用小波变换分解成低频和高频部分,低频的小波系数平均值作为融合后的低频系数,高频细节系数根据不同区域特征选择方法以及对应输入图像小波系数的窗口区域方差来确定融合后高频小波系数,得到一个波段一幅图像.接着,将得到的图像再进行小波分解,采用低频图像的小波系数最小值作为融合后的低频系数,高频图像根据纹理一致性测度的纹理检测确定融合规则,用来调整高频小波系数,将来自不同图像的特征与细节融合在一起,并对融合图像质量进行了对比评价.实验结果表明,融合后的偏振图像不仅反映了场景的偏振信息,而且还包含了丰富的光谱信息,目标与背景的衬比度也得到了增强,为进一步的目标检测和识别提供了便利.     

融合改进梅尔谱特征和深信念网络的语音测谎算法

为实现非接触式谎言检测,特提出了以语谱特征为线索,结合深度学习的谎言检测方法。为提取谎言中微颤抖所引起的语谱局部能量变化,算法先对梅尔频谱进行了Hu矩处理,然后进行离散余弦变换去除相关性。该特征利用了Hu矩的正交不变性和平移不变性,能较好的体现出语谱中局部能量的集中方式。然后将所提取的特征作为改进深信念网络输入进行谎言识别。为提高受限玻尔兹曼机的并行回火训练算法中相邻温度链之间的交换率,训练算法先对Markov链的状态能量进行等能量的划分,使得每个能量环内的状态具有相似的能量,然后再进行交换以提高交换率从而优化整个网络的训练。在Columbia-SRI-Colorado数据库上的实验表明,谎言识别率达到了71.47%,比梅尔倒谱系数特征的识别率提高了3%,比传统的BayesNet分类算法提高了7%。      

全局特征及弱尺度融合策略的小样本语音情感识别

语音是一种短时平稳时频信号,因此大多数的研究者都通过分帧来提取情感特征。然而,分帧后提取的特征为局部特征,无法准确反应情感语音动态特性,故单纯采用局部特征往往无法构建鲁棒的情感识别系统。针对这个问题,先在不分帧的语音信号里通过多尺度最优小波包分解提取语句级全局特征,分帧后再提取384维的语句级局部特征,并利用Fisher准则进行降维,最后提出一种弱尺度融合策略来将这两种语句级特征进行融合,再利用SVM进行情感分类。基于柏林情感库的实验结果表明本文方法较单纯使用语句级局部特征最后识别率提高了4.2%到13.8%,特别在小样本的情况下,语音情感识别率波动较小。