Enhancement network: confining deep convolutional features based on hand-crafted rules

Despite convolutional neural network(CNN) is mature in many domains, the understanding of the directions where the parameters of the CNNs are learned towards, falls behind, and researches on the functions that the convolutional networks(ConvNets) learns are difficult to be explored. A method is proposed to guide ConvNets to learn towards the expected direction. First, for the sake of facilitating network converging, a novel feature enhancement framework, namely enhancement network(EN), is devised to learn parameters according to certain rules. Second, two types of hand-crafted rules, namely feature-sharpening(FS) and feature-amplifying(FA) are proposed to enable effective ENs, meanwhile are embedded into the CNN for the end-to-end learning. Specifically, the former is a tool sharpening convolutional features and the latter is the one amplifying convolutional features linearly. Both tools aim at the same spot achieving a stronger inductive bias and more straightforward loss functions. Finally, the experiments are conducted on the mixed National Institute of Standards and Technology(MNIST) and cooperative institute for Alaska research 10(CIFAR10) dataset. Experimental results demonstrate that ENs make a faster convergence by formulating hand-crafted rules.

Modified Cepstral Feature for Speech Anti-spoofing

The hidden danger of the automatic speaker verification(ASV)system is various spoofed speeches.These threats can be classified into two categories,namely logical access(LA)and physical access(PA).To improve identification capability of spoofed speech detection,this paper considers the research on features.Firstly,following the idea of modifying the constant-Q-based features,this work considered adding variance or mean to the constant-Q-based cepstral domain to obtain good performance.Secondly,linear frequency cepstral coefficients(LFCCs)performed comparably with constant-Q-based features.Finally,we proposed linear frequency variance-based cepstral coefficients(LVCCs)and linear frequency mean-based cepstral coefficients(LMCCs)for identification of speech spoofing.LVCCs and LMCCs could be attained by adding the frame variance or the mean to the log magnitude spectrum based on LFCC features.The proposed novel features were evaluated on ASVspoof 2019 datase.The experimental results show that compared with known hand-crafted features,LVCCs and LMCCs are more effective in resisting spoofed speech attack.

Detection of Lung Nodules on X-ray Using Transfer Learning and Manual Features

The well-established mortality rates due to lung cancers,scarcity of radiology experts and inter-observer variability underpin the dire need for robust and accurate computer aided diagnostics to provide a second opinion.To this end,we propose a feature grafting approach to classify lung cancer images from publicly available National Institute of Health(NIH)chest X-Ray dataset comprised of 30,805 unique patients.The performance of transfer learning with pre-trained VGG and Inception models is evaluated in comparison against manually extracted radiomics features added to convolutional neural network using custom layer.For classification with both approaches,Support VectorsMachines(SVM)are used.The results from the 5-fold cross validation report Area Under Curve(AUC)of 0.92 and accuracy of 96.87%in detecting lung nodules with the proposed method.This is a plausible improvement against the observed accuracy of transfer learning using Inception(79.87%).The specificity of allmethods is>99%,however,the sensitivity of the proposed method(97.24%)surpasses that of transfer learning approaches(<67%).Furthermore,it is observed that the true positive rate with SVM is highest at the same false-positive rate in experiments amongst Random Forests,Decision Trees,and K-Nearest Neighbor classifiers.Hence,the proposed approach can be used in clinical and research environments to provide second opinions very close to the experts’intuition.