This dataset supports the study of startle-related physiological responses during walking using a wireless smart assistive cane equipped with a Galvanic Skin Response (GSR) sensor. Data were collected from 22 healthy participants walking through university areas while exposed to sudden auditory stimuli designed to elicit startle responses. GSR, tonic, phasic and acceleration signals were recorded simultaneously using the smart cane and an Empatica EmbracePlus reference device. The dataset includes raw/preprocessed signals, extracted GSR features and statistical results from Linear Mixed-Effects Models (LMMs) assessing the effects of startle condition and motion. The analysis identified 12 of 20 extracted features as motion-robust on the smart cane, with moderate-to-large effect sizes (Cohen’s dz = 0.54–1.27). These results indicate that startle-related physiological information can be detected from a cane-embedded GSR sensor during walking despite motion artifacts, supporting its potential use in future assistive and wearable systems for ambulatory startle and stress monitoring.