The current public datasets of Pedestrian Attribute Recognition （PAR） have the characteristics of complicated attribute annotations and various collection scenarios， leading to the large variations of the pedestrian attributes in different datasets， so that it is hard to directly utilize the existing labeled information in the public datasets for PAR in practice. To address this issue， a multi-source and multi-label PAR method based on domain adaptation was proposed. Firstly， to transfer the styles of the different datasets into a unified one， the features of the samples were aligned by the domain adaption method. Then， a multi-attribute one-hot coding and weighting algorithm was proposed to align the labels with the common attribute in multiple datasets. Finally， the multi-label semi-supervised loss function was combined to perform joint training across datasets to improve the attribute recognition accuracy. The proposed feature alignment and label alignment algorithms were able to effectively solve the heterogeneity problem of attributes in multiple PAR datasets. Experimental results after aligning three pedestrian attribute datasets PETA， RAPv1 and RAPv2 with PA-100K dataset show that the proposed method improves the average accuracy by 1.22 percentage points， 1.62 percentage points and 1.53 percentage points respectively， compared to the method StrongBaseline， demonstrating that this method has a strong advantage in cross dataset PAR.

In big data environment, traditional malicious file detection methods have low detection accuracy for malicious files after code variant and confusion, and weak versatility of cross-platform malicious files. To resolve these problems, a malicious file detection method based on image texture and Convolutional Neural Network (CNN) was proposed. Firstly, a grayscale image generation algorithm was used to convert the executable files on Android and Windows platforms, namely.dex and.exe files, into corresponding grayscale images. Then, the texture features of these grayscale images were automatically extracted and learned by using CNN algorithm, to construct a malicious file detection model. Finally, a large number of unknown files were used to test the accuracy of the proposed model. The experimental results on a large number of malicious samples showed that the highest accuracy of the proposed model on Android platform and Windows platform reached 79.6% and 97.6%, and the average accuracy were approximately 79.3% and 96.8%, respectively. Compared with the texture fingerprint-based malicious code detection method, the accuracy of the proposed method was improved by about 20%. Experimenatal results indicate that the proposed method can effectively avoid the problems caused by manual screening features, greatly improve the detection accuracy and efficiency, successfully solve the cross-platform detection problem, and achieve an end-to-end malicious file detection model.

According to the continuous space model for emotion, an improved queuing voting algorithm was proposed to implement the fusion of multiple emotion classifiers for a good emotion recognition result. Based on hidden Markov model (HMM) and artificial neural network (ANN), three kinds of classifier were designed. Then, the improved queuing voting algorithm was used to fuse them. Experimental study had been carried out by using Mandarin emotional speech database recoded and emotional speech database respectively. The results prove that the improved queuing voting algorithm can attain better fusion effect than conventional fusion algorithm and excel any single classifier evidently. The provided algorithm is not only easy to implement, but also transplantable. It is suitable for the fusion of any emotion classifiers.