Based On Galvanic Skin Response Signal Emotional Real-time Identification And Adjustment Method Research In Human-computer Interaction Environment

Affective computing, as a frontier research topic, received wide attention of scholars both at home and abroad. Emotion recognition based on physiological signals has become an important research direction of affective computing, because of objective authenticity of the physiological signal. At the same time, with the improving of the human-computer interaction friendly demand, the importance of emotion regulation is growing. This article which aims at the emotion recognition of skin electric puts forward an improved K neighbor real-time recognition method. The results by use this approach has a significant promotion in both recognition effect and recognition efficiency. In terms of emotion regulation, the research referenced Gross and other scholar’s emotional adjustment model, putted forward human-computer interaction environment feasibility emotional adjustment model. And established open man-machine emotional interaction which can improve human-computer interaction harmony according to the theory of results. Specific work is as follows:(1) Making Galvanic Skin Response(GSR) signal acquisition experiment plan, establish Galvanic Skin Response emotional sample library. The research induced subjects’ joy, angry, sadness, fear four kinds of emotion via different emotions evoked video clips. And in the process of induced, Galvanic Skin Response signals were acquired.(2) Galvanic Skin Response signal preprocessing and feature extraction. This part of the work mainly includes signal interception, smoothing denoising, feature extraction, feature normalization, feature selection. Signal interception is cut the length of the signal for 15 seconds from a length of about 5 minutes of Galvanic Skin Response signal, the purpose is to find the signal fragments which most represent the participants’ induced emotions. Smooth denoising removed noise spectrum signals and got most real physiological signals using the wavelet denoising. In feature extraction process, the 30 features including average, maximum, minimum of first-order difference, and median of frequency domain were extracted. The characteristic value with different orders of magnitude, especially in the time domain characteristics and frequency domain characteristics. So the characteristic values normalization can make every feature has equal voting rights in classification. Feature selection is to use the Relief features selection algorithm to cut out six invalid features.(3) Put forward an improved K neighbor real-time recognition method. K neighbor algorithm mainly time-consuming is looking for test samples recently K training samples. So using K neighbor fast search algorithm change the original ways to look for training samples. At the same time, using real-time identification characteristic, the research adopts a half supervision K neighbor recognition method to let the current test sample of C before a sample in time series are also involved in recognition process. Research validations recognition rate and running time when C and K have different values, and compares the original K neighbor method, proved that the improved scheme effective in recognition rate and the running time.(4) Human-computer interaction environment feasibility emotion regulation model is set up. Gross’ emotion regulation process model can effectively response the emotion principle and adjust method of each stage, but don’t adapt to the man-machine interaction environment. This paper reference Gross emotion regulation process model and Bonanno sequence of emotional self-regulation model, established human-computer interaction environment feasibility emotional adjustment model.(5) Designed and implemented an open man-machine emotional interaction helper. Users in the use of this software, you can use an external equipment real-time acquisition skin electric signal, through the background algorithm operation real-time rendering emotional recognition result. When software tested the users in the negative emotion for a long time, it will be based on the characteristics of user customization push, including video, music, games and words, different emotion regulation scheme which make the human-computer interaction more harmonious and friendly.