Understanding the comprehensive sound perceptions is a major challenge in translating the affective needs and responses of drivers and passengers into the humancentered design of seat-belt warning sound (SBWS). Due to the various sound impressions of SBWS, sound quality has multiple impacts on driving safety, pleasure, stress, and so forth. This paper aims to derive a specified psychological factor structure for the sound perceptions of SBWS and to rebuild the connections between the psychological and acoustic attributes. Kansei (emotional or affective) evaluation is employed to collect the affective responses to 20 sound stimuli from 10 experts and 134 other participants. The perceived sound quality is expressed by 11 psychological attributes: "cheerful," "interesting," "lovely," "comfortable," "nervous," "cordial," "advanced," "crisp," "nostalgic," "aroused," and "weak." Russell's emotional state model is also employed to map the sounds into a Pleasure-Arousal Kansei space for further analysis. For acoustic attributes, we measure amplitude, frequency, pause ratio, repetition speed, and element number, which are easy to be tuned according to the design guidelines listed as follows: First, a decrease in sound frequency results in sleepiness perception (less perceived urgency); second, fewer element (peak) number per cycle results in displeasure perception (less satisfaction); third, lower repetition speed results in pleasure and sleepiness perceptions simultaneously. The multiple linear regression model further indicates the quantitative psychoacoustic relationships, which helps sound designers to identify the key acoustic attributes and predict the affective responses for the human-centered design of SBWS.