The work aims to explore the appropriate filtering methods and physiological characteristic indicators of watch products for specific groups of people. Based on the current status of research on physiological signal filtering and feature extraction in the emotional product design process, an experimental platform for emotional imagery of watches was built. Skin electrodermal signals, electrocardiogram signals and electroencephalogram signals were used to explore the differential associations between different filtering and feature extraction methods and emotional imagery under population characteristics (college students, university teachers and white-collar workers) based on Lasso regression. The skin electrodermal signal indicators showed low volatility under various filtering methods, the electrocardiogram signal indicators were greatly affected by the "high-pass + low-pass" filtering method, and the electroencephalogram signal indicators were greatly affected by the "IMF + wavelet transform" filtering method. University teachers showed higher stability in filtering methods compared with college students and white-collar workers. The dimensions and number of physiological emotional features retained in the Lasso regression diagrams of valence and arousal were different for different groups of people. In the emotional product design, the corresponding filtering methods and feature extraction indicators should be optimized according to the physiological characteristics of different groups of people, so as to accurately establish the "physiological-product emotional imagery" mapping model and realize the effective integration of product form and emotional state.