A 3D Geometry Model of Vocal Tract Based on Smart Internet of Things

The Internet of Things(IoT)plays an essential role in the current and future generations of information,network,and communication development and applications.This research focuses on vocal tract visualization and modeling,which are critical issues in realizing inner vocal tract animation.That is applied in many fields,such as speech training,speech therapy,speech analysis and other speech production-related applications.This work constructed a geometric model by observation of Magnetic Resonance Imaging data,providing a new method to annotate and construct 3D vocal tract organs.The proposed method has two advantages compared with previous methods.Firstly it has a uniform construction protocol for all speech organs.Secondly,this method can build correspondent feature points between different speech organs.There are less than three control parameters can be used to describe every speech organ accurately,for which the accumulated contribution rate is more than 88%.By means of the reconfiguration,the model error is less than 1.0 mm.Regarding to the data from Chinese Magnetic resonance imaging(MRI),this is the first work of 3D vocal tract model.It will promote the theoretical research and development of the intelligent Internet of Things facing speech generation-related issues.

Profile of Patients with Stroke and Disorders of the Vocal Tract

Background: The present work aims to characterize the profile of patients with stroke treat at a hospital located in the Region of the Mata of Minas Gerais, Brazil, considering the findings of the clinical vocal tract, kind of stroke, age and gender of such patients. Methodology: To obtain data, the clinical profile of 133 patients with a clinical or tomography diagnosis of stroke was analyzed, and the results were presented in percentage. For quantitative data average and analysis the tests were done with associations that held χ2 test, and for significance it was considered p Results: From the total of patients, 63 were women, accounting for 47.4% and the other 52.6% were males. Clinically, they were characterized with the highest percentage for ischemic stroke (89.4%) compared to the hemorrhagic type (10.6%). Most of them were referred for computed tomography (86.5%) and remained hospitalized for an average of 6.496 ± 7.372 days. Similar percentages were obtained in the analysis of the population in question, when considering if they had (54.1%) or not (49.6%) any damage in their speech, language skills or swallowing. There were different types of disabilities in patients with stroke. Men with an average age of 69.8 ± 13.9 presents mostly ischemic stroke, and the majority of patients with stroke had hemiplegia and abnormalities of the vocal tract, dysphasia, and aphasia. While older patients had an ischemic stroke and were presented with left hemiplegia, the younger ones suffered from hemorrhagic strokes that caused a disability characterized as right hemiplegic. Conclusion: Our results show important conclusions regarding the clinical evolution of the vocal tract of patients who suffered strokes during the period of the analysis, being useful for better comprehension of how the vocal tract from these patients evolved according to the kind of stroke, sex and age also allowing a contraposition with other future statistics periods available in literature. It can also be pointed out the difficulties in diagnosing the stroke and the concern with the immediate care, but not with its continuance or with its multidisciplinary approach, giving an evident life risk through dysphasia and the increase of permanent damage when there isn’t an appropriate work done with the patients.

Research of whispered speech vocal tract system conversion based on universal background model and effective Gaussian components

Directing to the weakness of the present fixed values mapping methods （method_F）, a vocal tract system conversion method based on the universal background model （UBM） is proposed for improving the performance of the speech conversion system from Chinese whis- pered speech to normal speech. For the numerous components of UBM, the errors produced by the acoustical probability density statistical model can＇t be ignored. Thus an effective Gaus- sian mixture components chosen method based on the posterior probability summation of the minimum spectral distortion is developed to optimizing the system performance. The proposed method （method_U） is analyzed and compared using the performance index （PI） based on Itakura-Saito spectral distortion measure. It is shown experimentally that the performance of method_U is more stability for different speakers and different phonemes than that of method_F. The average PI of method_U is better than method_F. It is shown that by selecting effective Gaussian mixture components, the PI of method_U can be further improved 5.11%. Subjective auditory tests also show that the proposed method can improve the definition and intelligibility of conversion speech.

A new speech synthesis method based on the LMA vocal tract model

A new speech synthesis algorithm based on the LMA filter in Chinese text-to-speech systern is introduced. Using this method, the system can not only generate speech with higher quality, but also have a more powerful ability to modify the prosodic parameters, which ensures a far more natural and intelligible synthesized speech than ever before. First, the fundamental principles of the LMA filter and the construction of the synthesizer are presented, then, how to modify the acoustic parameters with this synthesizer is described; finally, the quantitative evaluation of the system's performance is shown while compared with a relatively successful PSOLA synthesizer KDTALK_1

超声波水表的原理以及采样频率对计量性能影响的探讨

超声波水表作为新一代兴起的智能水表,因其计量准确度高、流量范围宽、维护简单以及更加符合用水管理的智能化、自动化需求等优势,逐渐应用于生产生活用水的各个方面。该文详细了介绍超声波水表的工作原理和结构组成,并通过试验解释和验证了采样频率对计量性能的影响。

Simulation of Human Phonation with Vocal Nodules

The geometric and biomechanical properties of the larynx strongly influence voice quality and efficiency. A physical understanding of phonation natures in pathological conditions is important for predictions of how voice disorders can be treated using therapy and rehabilitation. Here, we present a continuum-based numerical model of phonation that considers complex fluid-structure interactions occurring in the airway. This model considers a three-dimensional geometry of vocal folds, muscle contractions, and viscoelastic properties to provide a realistic framework of phonation. The vocal fold motion is coupled to an unsteady compressible respiratory flow, allowing numerical simulations of normal and diseased phonations to derive clear relationships between actual laryngeal structures and model parameters such as muscle activity. As a pilot analysis of diseased phonation, we model vocal nodules, the mass lesions that can appear bilaterally on both sides of the vocal folds. Comparison of simulations with and without the nodules demonstrates how the lesions affect vocal fold motion, consequently restricting voice quality. Furthermore, we found that the minimum lung pressure required for voice production increases as nodules move closer to the center of the vocal fold. Thus, simulations using the developed model may provide essential insight into complex phonation phenomena and further elucidate the etiologic mechanisms of voice disorders.

Some Tonal and Rhythmical Sequences in the Vocal Language of Dogs as Significant Earthquake Precursors

A monitoring of multiple physical parameters in a moderate seismic area in Western Piedmont (NW Italy) and the simultaneous observation of the behaviour of numerous species of domestic and wild animals gave in a period of over twenty years the possibility to distinguish the unusual animal behaviours due to local earthquake nucleation from other causes. In particular, the observation of the body and vocal language of dogs (Canis familiaris) in the same area has permitted not only to specify the different meanings of vocal language in connection to their body language, but also to classify the minimum elements into a vocal language that is linked together by tonal and rhythmical sequences of sounds that form a semantic lexicon. The usage of the same tonal and rhythmical vocal sequences in similar or identical situations, which are experienced by different groups of dogs, induces us to verify whether it could be possible to link particular vocal sequences to precise physical anomalies before earthquakes. The individuation of physical anomalies due to an earthquake nucleation or due to a hydro-geological destabilization, is possible thanks to a continuous long-term monitoring of some parameters. Moreover, the complexity of the vocal language of dogs increases if the dogs live in an area with a law population density. Then the correlation between some vocal sequences and some seismic precursors is better if dogs live free in yard or on farms, if they are in good health, and if they can establish a strong social relation of group. When dogs live closed in yards of houses that are far apart, they communicate with each other with an amazing vocal language, full of questions and answers, imitations of sequences, and information about situations that may be harmful to them.

基于听觉特性和发声特性的语种识别

针对现有的方法在低信噪比环境下语种识别性能不佳,提出了一种耳蜗滤波系数和声道冲激响应频谱参数相互融合的语种识别方法.该方法表征了人的耳蜗听觉特性和发声特性,首先提取模拟人耳听觉特性的耳蜗滤波系数,再融合表征人的发声特性的声道冲激响应频谱参数,最后采用高斯混合通用背景模型对所提方法在语种识别上进行测试.实验结果表明,在4种信噪比环境下,该方法优于其他对比方法;相对于基于深度学习的对数Mel尺度滤波器能量特征,识别正确率提升了16.1%,与其他方法相比有较大程度的提升.

基于声门波和声道特征的语音情感识别

语音情感识别是实现自然人机交互不可缺失的部分,是人工智能的重要组成部分。发音器官的调控引起情感语音声学特征的差异,从而被感知到不同的情感。传统的语音情感识别只是针对语音信号中的声学特征或听觉特征进行情感分类,忽略了声门波和声道等发音特征对情感感知的重要作用。在我们前期工作中,理论分析了声门波和声道形状对感知情感的重要影响,但未将声门波与声道特征用于语音情感识别。因此,本文从语音生成的角度重新探讨了声门波与声道特征对语音情感识别的可能性,提出一种基于源-滤波器模型的声门波和声道特征语音情感识别方法。首先,利用Liljencrants-Fant和Auto-Regressive eXogenous(ARX-LF)模型从语音信号中分离出情感语音的声门波和声道特征;然后,将分离出的声门波和声道特征送入双向门控循环单元(BiGRU)进行情感识别分类任务。在公开的情感数据集IEMOCAP上进行了情感识别验证,实验结果证明了声门波和声道特征可以有效的区分情感,且情感识别性能优于一些传统特征。本文从发音相关的声门波与声道研究语音情感识别,为语音情感识别技术提供了一种新思路。

基于双因子高斯过程动态模型的声道谱转换方法

针对作者已经提出的双因子高斯过程隐变量模型(Two-factorGaussianprocesslatentvariablemodel,TF-GPLVM)用于语音转换时未考虑语音的动态特征,并且模型训练时需要估计的参数较多的问题,提出引入隐马尔科夫模型(Hidden Markov model,HMM)对语音动态特征进行建模,并利用HMM隐状态对各帧语音进行关于语义内容的概率软分类,建立了分离精度更高、运算负荷较小的双因子高斯过程动态模型(Two-factor Gaussian process dynamic model,TF-GPDM).基于此模型,设计了一种全新的基于说话人特征替换的语音声道谱转换方案.主、客观实验结果表明,无论是与传统的统计映射和频率弯折转换方法相比,还是与双因子高斯过程隐变量模型方法相比,本文方法都获得了语音质量和转换相似度的提升,以及两项性能的更佳平衡.

针对语音变换的语音篡改检测

针对使用语音变换技术的语音篡改,提出一种自动检测方法。在分析语音变换基本模型和变换语音失真的基础上,提取语音信号的声道参数以及相关的信号统计量,并通过支持向量机递归特征消除法,选择出对语音变换比较敏感的特征作为分类特征,使用支持向量机进行语音变换检测和变换语音的说话人性别判别。对于一种语音变换软件的实验结果表明,该方法具有较高的检测准确率,其中语音变换检测的平均准确率为94.90%,变换语音的说话人性别判别平均准确率为92.09%。

语声转换技术发展及展望

语声转换通过改变语音信号的声学特征参数来调整语音的个性特征,从而使得转换后的源说话人语音听起来就像是目标说话人的声音一样。系统地介绍了当前语声转换技术的发展状况,在描述语声转换技术的应用场景和系统框架的基础上,着重阐述了系统的转换模块,即声道特性的转换和韵律转换,特别是重点介绍了声道特性的转换算法。简要地介绍了系统性能的测试方法,最后对全文进行了总结,并针对当前语声转换技术还存在的一些问题,对未来的发展进行了展望。

鸣禽白腰文鸟前脑古纹状体粗核性双态发育的神经机制

对鸣禽白腰文鸟 (Lonchurastriata)发声控制核团古纹状体粗核 (robustnucleusofarchistriatum ,RA)的性双态分化过程进行了组织学研究 ,并应用双向神经示踪剂 (biotinylateddextranamine ,BDA) ,追踪新纹状体外侧巨细胞核 (lateralnucleusmagnocellularisofanteriorneostriatum ,LMAN)和高级发声中枢 (highvocalcenter,HVC)与RA建立纤维联系的时间和过程。结果发现 :5～ 3 5日龄段为雌雄RA体积、神经元大小和神经元密度变化最集中的时间。在该时段内 ,RA体积、神经元大小均增加 3～ 4倍 ,而RA神经元密度减少约 4倍。这些变化在雌雄间无显著差异 (P >0 0 5 ,非配对 ,双尾t 检验 ) ,但与RA同LMAN、HVC建立神经联系的时间一致。RA同LMAN、HVC建立联系的时间分别为 5～ 15和 15～ 3 5日龄。 4 5日龄后 ,RA体积大小在雌、雄间出现显著差别 (P <0 0 5 )。 4 5～ 60日龄为雌鸟神经元凋亡数量最多时期 ,4 5和 60日龄神经元凋亡数分别为 19 4± 8 0和 17 9± 8 2 (× 10 3/mm3)。结果提示 :4 5日龄后雌雄鸟RA体积和神经元凋亡的变化可能是鸣禽发声核团性双态产生的主要原因。

普通话播音员发音共鸣效果的声学特性分析

为研究普通话播音员的发音共鸣特性及其与普通人的差异,本文采集了32名播音员和37名普通人发音时的语音信号和喉部皮肤振动信号,测量了基频f0、等效声压级SPLeq和皮肤振动幅值SAL等参数,并计算了共鸣放大比例系数。结果表明,无论是自然还是大声说话状态下,男女播音员均可以用较小的嗓音能量,产生更大的音量输出,说明播音员具有更好的发音共鸣放大效果。其放大增益比普通人约高3dB,相当于两个普通人同时说话音量叠加的效果。

基于加权全局时频特征的易混淆词识别

针对易混淆词特征差异小、分类决策困难的特点，提出了一种新的语音识别特征．该特征可以根据待识单词的发音特点，通过选用合适的基函数及加权处理，突出混淆单词特征之间的差异性；同时，根据其矢量维数相等的特点，利用静态神经网络分类决策能力强、容错性好的优点进一步提高系统的识别性能．实验结果表明，所用方法比传统的ＤＨＭＭ方法和其他神经网络语音识别方法具有更好的识别效率．

基于非齐次隐马尔可夫模型的特定人元音的识别方法

针对特定人汉语元音的语音识别,提出一种基于非齐次隐马尔可夫模型的识别方法.该方法首先提取声道频率响应作为特征参数,然后建立非齐次隐马尔可夫模型来更为精确地刻画真实的语音现象,接着进行语音识别实验,并与齐次隐马尔可夫模型进行比较.实验结果表明该方法可以使特定人的元音的识别率达到98.73%,明显改变了识别系统的性能.该方法具有很好的理论研究前景和实际应用价值.

针对构音异常辅助治疗的声道仿真研究

针对构音异常,本文提出了使用声道仿真来实现辅助治疗的方法。基于声道是一个弯曲的、三维的具有慢时变特性的声学管道,并且在声道中的声波传播是平面波的特性,可以把声道等效于一个具有不同截面的圆柱体或者椭圆体管道。使用极点形式,在牛顿插值的基础上得到共振峰。对声道进行了60段分段,通过经验公式得到声道在不同部位的面积。定义了描述声道特性的9个参数,进而对这9个参数使用Corana算法进行优化。使用辐射模型描述声音从嘴唇辐射出去以后的特性。最后进行声音的合成,这个声音可用于反馈治疗。经过实验证明,这种声道仿真模型可以为制定合适治疗方法提供参考。

浊音短时复倒谱的新模型及用于语音同态解卷积的研究

本文介绍一种浊音短时复倒谱的新模型，通过与传统模型的分析对比，提出了将该模型用于同态解卷积恢复声道冲激响应的新方法，并进行了实验研究。对解卷积中的问题进行了分析探讨。

基于语音质量参数的语音传输信道检测算法

针对不同编码算法对语音质量的影响,采用语音质量改进参数构造特征矢量,提出了一种基于SVM多级判决的语音编码检测算法,实现了对不同信道的识别。采用统计方法对这组参数的区分性进行了分析,在此基础上设计了一种高效的信道检测方案,结合实际数据,测试了算法性能,并分析了语音长度对性能的影响。实验结果表明,该算法能有效地提高信道检测准确率。

基于MATLAB的LSF滤波器实现

线谱对参数具有优越的量化特性和内插特性,以线谱对参数作为滤波器系数的LSF滤波器是一种适用于高压缩率声码器的声道滤波器。MATALAB功能强大且编程方便,广泛应用于各种算法的性能评估。文章简要介绍了LSF声道滤波器原理,给出了用MATLAB实现LSF滤波器的流程图和程序代码。最后设计了一个LSF声道滤波器的应用实例,并对应用实例进行了MATLAB仿真,给出了应用实例的仿真结果。