An end-to-end model with angular interval embedding was constructed to solve the problems of complicated multiple steps and weak generalization ability in the traditional voiceprint recognition model based on the combination of identity vector (i-vector) and Probabilistic Linear Discriminant Analysis (PLDA). A deep convolutional neural network was specially designed to extract deep speaker embedding from the acoustic features of voice data. The Angular Softmax (A-Softmax), which is based on angular improvement, was employed as the loss function to keep the angular interval between the different classes of features learned by the model and make the clustering of the similar features closer in the angle space. Compared with the method combining i-vector and PLDA, it shows that the proposed model has the identification accuracy of Top-1 and Top-5 increased by 58.9% and 30% respectively and has the minimum detection cost and equal error rate reduced by 47.9% and 45.3% respectively for speaker verification on the public dataset VoxCeleb2. The results verify that the proposed end-to-end model is more suitable for learning class-discriminating features from multi-channel and large-scale datasets.

Concerning the poor robustness and low hiding strength of existing file hiding method based on Universal Serial Bus (USB), a new file hiding method based on capacity disguised and double file system was proposed. By analyzing the characteristics and management mechanism of Nand flash chips, the capacity disguise was achieved to deceive the host by tampering equipment capacity value in Command Status Wrap (CSW). Based on the memory management mechanism of the Flash Translation Layer (FTL), the storage area was divided into two parts including the hiding area and the common area by different marks, and a double file system was established using format function. Request for switching file system was sent by writing specific data, then it was achieved after user authentication to realize secure access to hiding areas. The experimental results and theoretical analysis show that the proposed method can achieve hiding file which is transparent to operating system, moreover, it is not affected by device operation and has better robustness and stronger hiding effect with respect to the methods based on hooking Application Programming Interface (API), modifying File Allocation Table (FAT) or encryption.

This thesis focused on the traditional and message-based SoftMan's communication approach which has some problems in the aspects of expression ability, communication efficiency and quality. Based on the early research in SoftMan system and its communication theory, as well as SoftMan cogmatics model and context awareness mechanism, this thesis proposed the Context driven SoftMan Knowledge Communication (CSMKC) framework by learning from mature Agent communication language specification. First, the message layer, the knowledge layer and the scenarios layer in knowledge communication framework were designed; second, from the three aspects of implementation of message layer, knowledge layer and scenarios layer, the key points of knowledge communication achievements of scenario-driven SoftMan were introduced; finally, different SoftMan's communication in knowledge grade and the maintenance of scenario context were realized basically. The experimental results show that when the later content has high dependence on communication scenario, compared with the traditional message-based SoftMan communication approach, the communication overhead per unit time of CSMKC reduces by 46.15% averagely. Thus, the higher dependence on the scene, the more obvious CSMKC advantages in terms of reducing communication while accomplishing a task in the system.