基于动态卷积递归神经网络的语音情感识别

动态情感特征是说话人独立语音情感识别中的重要特征。由于缺乏对语音中时频信息的充分挖掘，现有动态情感特征表征能力有限。为更好地提取语音中的动态情感特征，提出一种动态卷积递归神经网络语音情感识别模型。基于动态卷积理论构建一种动态卷积神经网络提取语谱图中的全局动态情感信息，使用注意力机制分别从时间和频率维度对特征图关键情感区域进行强化表示，同时利用双向长短期记忆网络对谱图进行逐帧学习，提取动态帧级特征及情感的时序依赖关系。在此基础上，利用最大密度散度损失对齐新个体特征与训练集特征分布，降低个体差异性对特征分布产生的影响，提升模型表征能力。实验结果表明，该模型在CASIA中文情感语料库、Emo-db德文情感语料库及IEMOCAP英文情感语料库上分别取得59.50%、88.01%及66.90%的加权平均精度，相较HuWSF、CB-SER、RNN-Att等其他主流模型识别精度分别提升1.25～16.00、0.71～2.26及2.16～8.10个百分点，验证了所提模型的有效性。     

基于知识蒸馏和生成对抗网络的远场语音识别

为了进一步利用近场语音数据来提高远场语音识别的性能,提出一种基于知识蒸馏和生成对抗网络相结合的远场语音识别算法.该方法引入多任务学习框架,在进行声学建模的同时对远场语音特征进行增强.为了提高声学建模能力,使用近场语音的声学模型（老师模型）来指导远场语音的声学模型（学生模型）进行训练.通过最小化相对熵使得学生模型的后验概率分布逼近老师模型.为了提升特征增强的效果,加入鉴别网络来进行对抗训练,从而使得最终增强后的特征分布更逼近近场特征.AMI数据集上的实验结果表明,该算法的平均词错误率（WER）与基线相比在单通道的情况下,在没有说话人交叠和有说话人交叠时分别相对下降5.6%和4.7%.在多通道的情况下,在没有说话人交叠和有说话人交叠时分别相对下降6.2%和4.1%.TIMIT数据集上的实验结果表明,该算法获得了相对7.2%的平均词错误率下降.为了更好地展示生成对抗网络对语音增强的作用,对增强后的特征进行了可视化分析,进一步验证了该方法的有效性.     

基于尺度相关-双向长短期记忆网络模型的说话人识别

说话人识别通过语音对说话人进行身份认证,然而大部分语音在时域与频域具有分布多样性,目前说话人识别中的卷积神经网络深度学习模型普遍使用单一的卷积核进行特征提取,无法提取尺度相关特征及时频域特征。针对这一问题,提出一种尺度相关卷积神经网络-双向长短期记忆(SCCNN-BiLSTM)网络模型用于说话人识别。通过尺度相关卷积神经网络在每一层特征抽象过程中调整感受野大小,捕获由尺度相关块组成的尺度特征信息,同时引入双向长短期记忆网络保留与学习语音数据的多尺度特征信息,并在最大程度上提取时频域特征的上下文信息。实验结果表明,SCCNN-BiLSTM网络模型在LibriSpeech和AISHELL-1数据集上迭代50 000次时的等错率为7.21%和6.55%,相比于ResCNN基线网络模型提升了25.3%和41.0%。     

基于段级特征主成分分析的说话人识别算法

为了提高说话人识别(SR)系统的运算速度,增强其鲁棒性,以现有的帧级语音特征为基础,提出了一种基于段级特征主成分分析的说话人识别算法。该算法在训练和识别阶段以段级特征代替帧级特征,然后用主成分分析方法对段级特征进行降维、去相关。实验结果表明,该算法的系统训练时间、测试时间分别为基线系统的47.8%、40.0%,同时识别率略有提高,抑制了噪声对说话人识别系统的影响。该结果验证了基于段级特征主成分分析的说话人识别算法在识别率有所提高的情况下取得了较快的识别速度,同时在不同噪声环境下的不同信噪比情况下均可以提高系统识别率。     

噪声环境下基于特征信息融合的说话人识别

针对在干净的语音环境下说话人识别率很高,但噪声环境下说话人识别率急剧下降的问题,提出了一种在噪声环境下,利用信噪比权重对说话人的特征信息MFCC系数和基音周期进行非线性融合,同时对MFCC特征参数进行基于帧信噪比权重得分,并同传统的高斯混合模型算法和基于FO-MFCC联合分布的特征融合方法,在噪声环境下分别进行了说话人识别的性能比较,同时对提出的融合算法进行了仿真实现.实验结果表明:在噪声的环境下方法相比上述传统说话人识别方法,性能有了明显的提高,在干净的语音环境下性能相当.     

基于PCANN的说话人识别方法研究

本文利用主分量分析神经网络(PCANN)方法,得到一种新的说话人语音特征。该特征通过对相继几帧语音特征参数组成的特征向量作主分量分析得到.新的特征能有效的引入帧间相关信息,减小冗余度,削弱噪声的影响。实验表明,新特征提高了系统的识别性能。     

基于DBM-LSTM的多特征语音情感识别

为增强不同情感特征的融合程度和语音情感识别模型的鲁棒性,提出一种神经网络结构DBM-LSTM用于语音情感识别。利用深度受限玻尔兹曼机的特征重构原理将不同的情感特征进行融合;利用长短时记忆单元对短时特征进行长时建模,增强语音情感识别模型的鲁棒性;在柏林情感语音数据库上进行分类实验。研究结果表明,与传统识别模型相比,DBM-LSTM网络结构更适用于多特征语音情感识别任务,最优识别结果提升11%。     

基于渐进比率掩蔽目标的自适应噪声估计方法

基于深度学习的语音增强算法的性能通常优于传统的基于噪声抑制的语音增强算法。然而当训练数据和测试数据之间存在不匹配时,基于深度学习的语音增强算法通常无法正常工作。针对上述问题,提出一种新的基于渐进比率掩蔽（PRM）的自适应噪声估计（PRM-ANE）方法,并把它作为语音识别系统的预处理方法。所提方法综合利用了具有帧级别的噪声跟踪能力的改进最小统计量控制递归平均（IMCRA）算法和具有学习噪声和语音之间复杂非线性映射关系的渐进学习算法这两种算法。首先,使用二维卷积神经网络（2D-CNN）学习随信噪比（SNR）增加的PRM;其次,通过传统的帧级语音增强算法组合句子级估计的PRM,进行语音增强;最后,将基于多级别信息融合的增强语音直接作为语音识别系统的输入,从而提高识别系统性能。在CHiME-4真实测试集上的实验结果表明,所提方法可以实现7.42%的相对字识别错误率（WER）,与IMCRA语音增强方法相比下降了51.41%,可见所提方法能够有效提升下游识别任务的性能。     

语音识别中的情感信息分析

在语音识别和语音合成的技术领域内,对说话人情绪情感的识别以及在语音合成中再现某些情绪情感都非常重要。对语音的感知其实是人们利用各种感觉器官同时接受各种形式信息的结果,如何在人机交互中有效地利用各种形式的语音信息以达到最佳的信息传递效果,是今后语音信息处理研究的发展方向。     

语音情感识别研究综述

语音不仅包含说话人所要表达的语义信息,也蕴含着说话人所要表达的情感信息.语音情感识别是人机情感交互的关键,对语音情感的有效识别能够提升语音可懂度,使各种智能设备最大限度理解用户意图,提高机器人性化水平,从而更好地为人类服务.采用文献研究法从语音情感语料库、语音情感特征提取、语音情感模型的构建以及语音情感识别的应用等方面对其研究现状和进展进行了综述;同时,对其未来发展趋势也进行了展望.旨在尽可能全面地对语音情感识别技术进行详细分析,为相关研究人员提供有价值的学术参考.     

文本无关说话人识别中句级特征提取方法研究综述EI北大核心CSCD

句级(Utterance-level)特征提取是文本无关说话人识别领域中的重要研究方向之一.与只能刻画短时语音特性的帧级(Frame-level)特征相比,句级特征中包含了更丰富的说话人个性信息;且不同时长语音的句级特征均具有固定维度,更便于与大多数常用的模式识别方法相结合.近年来,句级特征提取的研究取得了很大的进展,鉴于其在说话人识别中的重要地位,本文对近期具有代表性的句级特征提取方法与技术进行整理与综述,并分别从前端处理、基于任务分段式与驱动式策略的特征提取方法,以及后端处理等方面进行论述,最后对未来的研究趋势展开探讨与分析.     

Object-based feature selection using class-pair separability for high-resolution image classification

ABSTRACT

With the increasing popularity of object-based image analysis (OBIA) since 2006, numerous classification and mapping tasks were reported to benefit from this evolving paradigm. In these studies, segments are firstly created, followed by classification based on segment-level information. However, the feature space formed by segment-level feature variables can be very large and complex, posing challenges to obtaining satisfactory classification performance. Accordingly, this work attempts to develop a new feature selection approach for segment-level features. Based on the principle of class-pair separability, the segment-level features are grouped according to their types. For each group, the contribution of each segment-level feature to the separation of a pair of classes is quantified. With the information of all feature groups and class pairs, the separability ranking and appearance frequency are considered to compute importance score for each feature. Higher importance score means larger appropriateness to select a feature. By using two Gaofen-2 multi-spectral images, the proposed method is validated. The experimental results show the advantages of the proposed technique over some state-of-the-art feature selection approaches: (1) it can better reduce the number of segment-level features and effectively avoid redundant information; (2) the feature subset obtained by the proposed scheme has good potential to improve classification accuracy.     

语音任务下声学特征提取综述

语音是一种重要的信息资源传递与交流方式,人们经常使用语音作为交流信息的媒介,在语音的声学信号中包含大量的说话者信息、语义信息和丰富的情感信息,因此形成了解决语音学任务的3个不同方向,即声纹识别(Speaker Recognition,SR)、语音识别(Auto Speech Recognition,ASR)和情感识别(Speech Emotion Recognition,SER),3个任务均在各自的领域使用不同的技术与特定的方法进行信息提取与模型设计。文中首先综述了3个任务在国内外早期的发展历史路线,将语音任务的发展归纳为4个不同阶段,同时总结了3个语音学任务在特征提取时所采用的公共语音学特征,并针对每类特征的侧重点进行了说明。然后,随着近年来深度学习技术在各个领域中的广泛应用,语音任务也得到了很好的发展,文中针对目前流行的深度学习模型在声学建模中的应用分别进行了分析,按照有监督、无监督的方式总结了针对3种不同语音任务的声学特征提取方式及技术路线,还总结了基于多通道并融合注意力机制的模型,用于语音的特征提取。为了同时完成语音识别、声纹识别和情感识别任务,针对声学信号的个性化特征提出了一个基于多任务的Tandem模型;此外,提出了一个多通道协作网络模型,利用这种设计思路可以提升多任务特征提取的准确度。     

Stratified pooling based deep convolutional neural networks for human action recognition

Video based human action recognition is an active and challenging topic in computer vision. Over the last few years, deep convolutional neural networks (CNN) has become the most popular method and achieved the state-of-the-art performance on several datasets, such as HMDB-51 and UCF-101. Since each video has a various number of frame-level features, how to combine these features to acquire good video-level feature becomes a challenging task. Therefore, this paper proposed a novel action recognition method named stratified pooling, which is based on deep convolutional neural networks (SP-CNN). The process is mainly composed of five parts: (i) fine-tuning a pre-trained CNN on the target dataset, (ii) frame-level features extraction; (iii) the principal component analysis (PCA) method for feature dimensionality reduction; (iv) stratified pooling frame-level features to get video-level feature; and (v) SVM for multiclass classification. Finally, the experimental results conducted on HMDB-51 and UCF-101 datasets show that the proposed method outperforms the state-of-the-art.     

Mismatched feature detection with finer granularity for emotional speaker recognition

The shapes of speakers＇ vocal organs change under their different emotional states, which leads to the deviation of the emotional acoustic space of short-time features from the neutral acoustic space and thereby the degradation of the speaker recognition performance. Features deviating greatly from the neutral acoustic space are considered as mismatched features, and they negatively affect speaker recognition systems. Emotion variation produces different feature deformations for different phonemes, so it is reasonable to build a finer model to detect mismatched features under each phoneme. However, given the difficulty of phoneme recognition, three sorts of acoustic class recognition--phoneme classes, Gaussian mixture model （GMM） tokenizer, and probabilistic GMM tokenizer--are proposed to replace phoneme recognition. We propose feature pruning and feature regulation methods to process the mismatched features to improve speaker recognition performance. As for the feature regulation method, a strategy of maximizing the between-class distance and minimizing the within-class distance is adopted to train the transformation matrix to regulate the mismatched features. Experiments conducted on the Mandarin affective speech corpus （MASC） show that our feature pruning and feature regulation methods increase the identification rate （IR） by 3.64% and 6.77%, compared with the baseline GMM-UBM （universal background model） algorithm. Also, corresponding IR increases of 2.09% and 3.32% can be obtained with our methods when applied to the state-of-the-art algorithm i-vector.     

Speaker identification using vowels features through a combined method of formants,wavelets, and neural network classifiers

This paper proposes a new method for speaker feature extraction based on Formants, Wavelet Entropy and Neural Networks denoted as FWENN. In the first stage, five formants and seven Shannon entropy wavelet packet are extracted from the speakers’ signals as the speaker feature vector. In the second stage, these 12 feature extraction coefficients are used as inputs to feed-forward neural networks. Probabilistic neural network is also proposed for comparison. In contrast to conventional speaker recognition methods that extract features from sentences (or words), the proposed method extracts the features from vowels. Advantages of using vowels include the ability to recognize speakers when only partially-recorded words are available. This may be useful for deaf-mute persons or when the recordings are damaged. Experimental results show that the proposed method succeeds in the speaker verification and identification tasks with high classification rate. This is accomplished with minimum amount of information, using only 12 coefficient features (i.e. vector length) and only one vowel signal, which is the major contribution of this work. The results are further compared to well-known classical algorithms for speaker recognition and are found to be superior.     

基于一致性图卷积模型的多模态对话情绪识别

多模态对话情绪识别是一项根据对话中话语的文本、语音、图像模态预测其情绪类别的任务。针对现有研究主要关注话语上下文的多模态特征提取和融合,而没有充分考虑每个说话人情绪特征利用的问题,提出一种基于一致性图卷积网络的多模态对话情绪识别模型。该模型首先构建了多模态特征学习和融合的图卷积网络,获得每条话语的上下文特征;在此基础上,以说话人在完整对话中的平均特征为一致性约束,使模型学习到更合理的话语特征,从而提高预测情绪类别的性能。在两个基准数据集IEMOCAP和MELD上与其他基线模型进行了比较,结果表明所提模型优于其他模型。此外,还通过消融实验验证了一致性约束和模型其他组成部分的有效性。     

结合多尺度特征学习与特征对齐的行人重识别

利用卷积神经网络对行人图像提取一个简单的全局特征,在复杂的行人重识别任务中无法获得令人满意的结果。局部特征学习的方式有助于获取更丰富的人体特征,但往往需要图像中的人体具有良好的空间对齐,而且,将人体各部分特征输入到独立的分支学习局部信息,忽略了人体各部分特征间的相关性,限制模型的性能提升。在此背景下,提出了一种新的多尺度特征学习算法,结合全局与局部特征学习得到更好的行人表示,提升复杂场景下模型的识别能力。对骨干网络不同深度输出的行人特征图,通过特征对齐模块对其执行空间变换,实现行人特征在空间上的矫正和对齐,进一步增强模型的泛化性能。在公开的大型行人重识别数据集上,与当前一些流行的方法进行了比较,验证了所提方法的有效性。     

Feature representation for facial expression recognition based on FACS and LBP

In expression recognition, feature representation is critical for successful recognition since it contains distinctive information of expressions. In this paper, a new approach for representing facial expression features is proposed with its objective to describe features in an effective and efficient way in order to improve the recognition performance. The method combines the facial action coding system(FACS) and "uniform" local binary patterns(LBP) to represent facial expression features from coarse to fine. The facial feature regions are extracted by active shape models(ASM) based on FACS to obtain the gray-level texture. Then, LBP is used to represent expression features for enhancing the discriminant. A facial expression recognition system is developed based on this feature extraction method by using K nearest neighborhood(K-NN) classifier to recognize facial expressions. Finally, experiments are carried out to evaluate this feature extraction method. The significance of removing the unrelated facial regions and enhancing the discrimination ability of expression features in the recognition process is indicated by the results, in addition to its convenience.     

T-STAM:基于双流时空注意力机制的端到端的动作识别模型

针对双流法进行视频动作识别时忽略特征通道间的相互联系、特征存在大量冗余的时空信息等问题,提出一种基于双流时空注意力机制的端到端的动作识别模型T-STAM,实现了对视频关键时空信息的充分利用。首先,将通道注意力机制引入到双流基础网络中,通过对特征通道间的依赖关系进行建模来校准通道信息,提高特征的表达能力。其次,提出一种基于CNN的时间注意力模型,使用较少的参数学习每帧的注意力得分,重点关注运动幅度明显的帧。同时提出一种多空间注意力模型,从不同角度计算每帧中各个位置的注意力得分,提取多个运动显著区域,并且对时空特征进行融合进一步增强视频的特征表示。最后,将融合后的特征输入到分类网络,按不同权重融合两流输出得到动作识别结果。在数据集HMDB51和UCF101上的实验结果表明T-STAM能有效地识别视频中的动作。