In the field of human interaction recognition, local features based on RGB video often cannot effectively distinguish approximate actions. The Depth image information and the color image information are merged in the recognition process, and a two⁃person interactive behavior recognition algorithm that integrates the depth information and the individual segmentation fusion is proposed.The algorithm firstly extracts the points of interest for RGB and Depth video, then uses 3DSIFT to describe the features on RGB video. The YOLO network is introduced into divide the left and right points of interest on the Depth video, and the local co⁃occurrence matrix is used for local correlation information description. Finally, the nearest neighbor classifier is used to classify the RGB features and Depth features, and further the recognition results are obtained by the decision⁃level fusion, which improves the accuracy of recognition. The results show that the combination of depth visual co⁃occurrence matrix can greatly improve the recognition accuracy of double interaction behavior, and the correct recognition rate of 90% of the actions in SBU Kinect interaction database can verify the effectiveness of the proposed algorithm.

With the popularization of network cameras and the continuous development of image processing technology, visual⁃based fire monitoring systems have received more and more attention. The current method of visual⁃based fire monitoring system is to detect the flame generated after a fire. The disadvantage is that the fire has already occurred during the alarm. The proposed fire detection system firstly finds out the suspected smoke area through the motion detection and the motion characteristics when the smoke just appears, and on this basis, the smoke and flame detection are carried out in the suspected smoke area. The smoke detection combines the smoke color characteristics and texture features. Flame detection uses the combination of flame two color features to improve detection accuracy. The experimental results show that the system can provide an alarm before the fire spreads and accurately detect the location of the fire ignition point, so the effectiveness of the proposed method is verified.

In order to solve the problem of large interference of environmental noise during speech recognition, an isolated word recognition algorithm based on empirical mode decomposition (EMD) and dynamic time warping (DTW) is proposed. In this method, the EMD algorithm is used to decompose the speech signal with poor performance into several basic mode functions (IMF) and remove the interference and noise in the original signal. Then, based on the DTW algorithm, the short-time zero crossing rate and short-time energy are used to detect the endpoint detection of speech signal. After the speech feature parameters are extracted, the speech signal is matched with the reference template. Finally, the shortest path between the reference template and the template to be measured is used as the recognition result. The simulation results show that the proposed algorithm can improve the recognition efficiency of speech and the accuracy of recognition.

Because of the rapid development of the convolutional neural network in recent years, the re-recognition of pedestrians has become another area of computer vision that is worthy of research after face recognition. The pedestrian re-recognition method involves verifying the similarity of two pedestrians or whether they are the same person and measuring the similarity of two features. And a pedestrian is a picture formed by multiple cameras, and its external conditions such as illumination, angle, and distance and it will increase the difficulty of verifying the similarities of two features. In this paper, a new method of pedestrian re-recognition based on combined Joint-Bayesian method and multi-camera shooting is proposed, which can effectively solve the problems caused by the change of shooting conditions of different cameras. Due to the good feature measurement ability of Joint-Bayesian, this paper studies a set of Joint-Bayesian matrices under different cameras, and combines with the global Joint Bayesian matrix to obtain a good recognition rate. On the Market-1501 and DukeMTMC-reID, the proposed method is used. The Rank-1 reached 88.09% and 80.07%, and the mAP reached 78.24% and 70.91%, which verified the effectiveness of the proposed method.

 

In order to improve the opacity of data collection, the complexity of each operating segment links and coarse tuning parameters manually in the process of data collection and analysis in the lad, a set of complete data collection and analysis system was designed by using data acquisition card of serial port and two channel, nonlinearity reset PSO algorithm and harmonious human-computer interaction interface. The working principles of frequency response data acquisition and data processing system, the composition structure and operation instructions were introduced. Equipment control and data acquisition were carried out after frequency instruction given by lower machine data acquisition system, the data was transmitted through the serial port communication to upper computer then through the Fourier transform into a data processing system. Test results showed that the system could collect data timely and accurately, display data in real time, process and save data automatically and precisely. The system was stable in testing process which guaranteed the accuracy of the data analysis.

In order to make the people find the appointed places more convenient in the unfamiliar environment, an intelligent guide system based on voice interaction in public places was designed and realized. First, according to the layout of the places, the shortest path could be gotten by Dijkstra algorithm. Then, the system could extract the feature of speech combine with the principle of speech recognition system, and match template based on Dynamic Time Warping algorithm. Finally, the people to arrive at the designated places could be guided by voice interaction. The system software and interface by MATLAB was operated, based on the actual tests, the designed system could be more intuitive and quick to implement the map navigation with speech interactive mode, and the voice recognition rate 87.75%.

The traditional sense of the flame size detection methods had the high detection error rate and easily influenced by sunlight conditions, and the flame region segmentation of images with complicated background still existed suspecting flame area. In order to improve the existing test methods， a method to detect the flame size puted forward in this paper ,based on RGB and HIS color space. This method combined two kinds of color space of flame characteristics, defining a new flame judgment constraint condition, and via the ratio between RGB color component values to eliminate light effects. By testing the images about different background, the results proved that the proposed method could effectively measure the flame size.