The present invention discloses a CACC (cooperative adaptive cruise control) driver limbs constant-velocity multichannel electromyographic signal feature extracting method, comprising the steps such as performing constant-velocity muscle strength operations of various magnitude operation degrees under various CACC conditions for a subject, acquiring signals and displacement and force feedback value information for a steering wheel, an accelerator pedal and a brake pedal, constructing a high-dimensional feature set, reducing dimensionality to obtain a low-dimensional feature set, and classifying the electromyographic signals. The extraction of CACC driver limbs constant-velocity multichannel electromyographic signal features can be effectively implemented.