Audio steganalysis based on reversed psychoacoustic model of human hearing

During the last decade, audio information hiding has attracted lots of attention due to its ability to provide a covert communication channel. On the other hand, various audio steganalysis schemes have been developed to detect the presence of any secret messages. Basically, audio steganography methods attempt to hide their messages in areas of time or frequency domains where human auditory system (HAS) does not perceive. Considering this fact, we propose a reliable audio steganalysis system based on the reversed Mel-frequency cepstral coefficients (R-MFCC) which aims to provide a model with maximum deviation from HAS model. Genetic algorithm is deployed to optimize dimension of the R-MFCC-based features. This will both speed up feature extraction and reduce the complexity of classification. The final decision is made by a trained support vector machine (SVM) to detect suspicious audio files. The proposed method achieves detection rates of 97.8% and 94.4% in the targeted (Steghide@1.563%) and universal scenarios. These results are respectively 17.3% and 20.8% higher than previous D2-MFCC based method.     

Source cell-phone recognition from recorded speech using non-speech segments

In a recent study, we have introduced the problem of identifying cell-phones using recorded speech and shown that speech signals convey information about the source device, making it possible to identify the source with some accuracy. In this paper, we consider recognizing source cell-phone microphones using non-speech segments of recorded speech. Taking an information-theoretic approach, we use Gaussian Mixture Model (GMM) trained with maximum mutual information (MMI) to represent device-specific features. Experimental results using Mel-frequency and linear frequency cepstral coefficients (MFCC and LFCC) show that features extracted from the non-speech segments of speech contain higher mutual information and yield higher recognition rates than those from speech portions or the whole utterance. Identification rate improves from 96.42% to 98.39% and equal error rate (EER) reduces from 1.20% to 0.47% when non-speech parts are used to extract features. Recognition results are provided with classical GMM trained both with maximum likelihood (ML) and maximum mutual information (MMI) criteria, as well as support vector machines (SVMs). Identification under additive noise case is also considered and it is shown that identification rates reduces dramatically in case of additive noise.     

Simulating Study of Dynamic Load Spectra Identification Method of Machinery in Cepstrum Domain

1 Introduction In general, data, curves and charts of random dynamic loads jointly are referred to as load spectra[1]. The load spectra reflect the laws of machine working loads, which play an effective role in the design and improvements of machines. Based on the theory of digital signal processing, this paper proposes a method for the identification of dynamic load spectra from the system of machine response signals. This is an adaptive windowing method based on the cep- strum domain of a dy…     

11/2维倒谱在滚动轴承故障诊断中的应用

从三阶累积量的概念出发,提出了1(1/2)维倒谱的概念,并从理论上推导出其能抑制噪声、提高信噪比.将其应用于实际的滚动轴承故障特征提取中,分析表明,与其他一般的信号处理方法相比1(1/2)维倒谱能有效地从复杂的信号成分中提取出特征信号来,且计算量小、精度高.     

基于提升小波变换的MFCC在目标识别中的应用

水下环境噪声对被动声呐目标的分类识别影响显著,为了提高水声信号识别系统对环境噪声的鲁棒性,采用提升小波变换方法提取MFCC特征,对水下被动声呐目标进行分类和识别。仿真实验表明,与传统MFCC和小波变换的MFCC相比,在信噪比相近的情况下,提升小波变换方法提取MFCC具有识别率较高、对噪声鲁棒性较好的显著优点。     

基于神经网络由语音预测视位参数

语音是由多个发音器官共同作用产生的，发音器官动作与语音之间有着内在的必然联系．研究了利用神经网络预测视位参数中的选择语音参数、确定输入语音时域范围、优化神经网络结构等因素．实验结果表明，线性预测参数加短时能量优于其他语音参数，前向协同发音较后向协同发音影响更大，反馈对前馈神经网络的性能有所改善．考虑到实验采用的是任意连续语流，均方误差约为0．0114的实验结果还是很有吸引力的．     

零相位同态反褶积

同态反褶积是一种利用复赛谱技术提取子波和进行子波反褶积的方法，它要求在频率域将子波相位正确地展开计算。在实际中做到这一点是十分困难的，为此本文提出零相位同态反褶积方法。在无井条件下，对于一般批量处理来说，提高数据的分辨率要比获得精确相位更为重要．在求取零相位子波时，除采用原有同态技术外，还采用了分形理论和自组织神经元方法进行聚类、提取子波，并获得了满意的结果。     

CNN: A speaker recognition system using a cascaded neural network

This work includes the design and implementation of both conventional, and neural network approaches to recognition of the speakers templates which are introduced to the system via a voice master card and preprocessed before extracting the features used in the recognltion. The conclusion is that the system performance in case of neural network is better than that of the conventional one, achieving a smooth degradation when dealing with nolsy patterns and higher performance when dealing with noise-free patterns.     

一种基于倒谱的人脸图像清晰度评价方法

针对人脸图像清晰度,给出了一种基于倒谱的定量评价方法。实验结果表明,与传统方法相比,所提出的清晰度评价方法能更加准确地评价出人脸图像的质量;采用该方法对人脸识别样本进行筛选,能有效地提高自动人脸识别的识别率,且对整体识别时间影响很小,因而更适合用作自动人脸识别中的图像质量评价。     

一种基于段级特征和自动标识的语言辨识算法

本文研究了一种结合"声学信息"和"音素配位学信息"进行语言辨识的新算法,首先在预处理中对语音进行自动分段,在特征层上引入带有长时信息的段级特征参数--段级移位差分倒谱,在模型层上利用高斯混合模型(Gaussi-an Mixture Model,GMM)将语音信号自动标识为符号序列,进而引入多元语言模型(Multi-gram Language Model,MLM)来对"音素配位学信息"进行建模,最后将"GMM得分"和"MLM得分"送入后端多分类支持向量机模型得到最终识别结果.相关实验表明,新系统不需手工标识的语料,识别速度快,对OGI标准语料库中的五种语言获得了开集正识率为78.84%的结果.