For problems such as low accuracy and poor real-time detection of existing radar non-contact vital signs detection， a human vital signs detection algorithm based on Frequency Modulated Continuous Wave （FMCW） radar was proposed. Firstly，the vital signs signal was obtained through the millimeter wave radar. Then， the adaptive decomposition and reconstruction of the vital signs signal were achieved using the improved Empirical Wavelet Transformation （EWT） algorithm. The best value of the spectrum division line was found by introducing Sparrow Search Algorithm （SSA） and Fuzzy Entropy （FE）. Finally，the heart rate and respiratory rate were calculated using the estimation algorithm with improved frequency interpolation. The superiority and robustness of the proposed algorithm were verified through comparative experiments with a medical critical care monitor. The experimental results showed that compared with Wavelet Transform （WT） algorithm， Complementary Ensemble Empirical Mode Decomposition （CEEMD） algorithm and Variational Mode Decomposition （VMD） algorithm， the Mean Square Error （MSE） was reduced by 77.65， 27.25 and 21.05， the Mean Absolute Percentage （MAPE） was reduced by 7.33， 4.33 and 3.42 percentage points， and the real-time performance was improved by 0.72 s， 16.74 s and 1.87 s. At the same time， the proposed algorithm also achieves the detection of Heart Rate Variability （HRV）.

An emotion recognition model based on Hybrid-Mel Gama Frequency Cross-attention Transformer modal （H-MGFCT） was proposed to address the issues of effectively mining single modal representation information and achieving full fusion of multimodal information in multimodal sentiment analysis. Firstly， Hybird-Mel Gama Frequency Cepstral Coefficient （H-MGFCC） was obtained by fusing Mel Frequency Cepstral Coefficient （MFCC） and Gammatone Frequency Cepstral Coefficient （GFCC）， as well as their first-order dynamic features， to solve the problem of speech emotional feature loss； secondly， a cross modal prediction model based on attention weight was used to filter out text features more relevant to speech features； subsequently， a Cross Self-Attention Transformer （CSA-Transformer） incorporating contrastive learning was used to fuse highly correlated cross modal information of text features and speech modal emotional features； finally， the cross modal information features containing text and speech were fused with the selected text features with low correlation to achieve information supplement. The experimental results show that the proposed model improves the accuracy by 2.83， 2.64， and 3.05 percentage points compared to the weighted Decision Level Fusion Text-audio （DLFT） model on the publicly available IEMOCAP （Interactive EMotional dyadic MOtion CAPture）， CMU-MOSI （CMU-Multimodal Opinion Emotion Intensity）， and CMU-MOSEI （CMU-Multimodal Opinion Sentiment Emotion Intensity） datasets， verifying the effectiveness of this model for emotion recognition.

The fine-grained image has characteristics of large intra-class variance and small inter-class variance， which makes Fine-Grained Image Categorization （FGIC） much more difficult than traditional image classification tasks. The application scenarios， task difficulties， algorithm development history and related common datasets of FGIC were described， and an overview of related algorithms was mainly presented. Classification methods based on local detection usually use operations of connection， summation and pooling， and the model training was complex and had many limitations in practical applications. Classification methods based on linear features simulated two neural pathways of human vision for recognition and localization respectively， and the classification effect is relatively better. Classification methods based on attention mechanism simulated the mechanism of human observation of external things， scanning the panorama first， and then locking the key attention area and forming the attention focus， and the classification effect was further improved. For the shortcomings of the current research， the next research directions of FGIC were proposed.

Considering the isolated cluster problem caused by traditional software modularization methods, a new metric named Improved Modularization Quality (IMQ) was proposed and used as the fitness function of an evolutionary algorithm to eliminate isolated clusters effectively. A mathematical programming model with the goal of maximizing IMQ was developed to represent software modularization problem. In addition, an Improved Genetic Algorithm (IGA) with competition and selection mechanism similarity was designed to solve this model. Firstly, a heuristic strategy based on edge contraction was used to generate high-quality solutions. Then the solutions were implanted as seeds into the initial population. At last, the proposed IGA was employed to further improve solution quality. Comparison experimental results prove that IMQ can effectively reduce the number of isolated clusters, and IGA has stronger robustness and ability of finding better solutions than Improved Hill Climbing Algorithm (IHC) and GA based on Group Number Encoding (GNE).

A combined image encryption algorithm was presented. The algorithm combined linear diffusion approach with Arnold mapped to disorder the image, and hyperchaotic sequences was used to encrypt the image. The mapped matrix and diffusion matrix were made up of chaotic sequences that were produced by the Logistic dynamic system, and they were used for disordering the color image to produce a disordered image on different bit planes to disorder the pixels. The disordered image was encrypted by means of the hyperchaotic sequences that were produced by the Chens system. The algorithm is simple and able to resist a variety of attacks, and can be easily implemented by hardware.