Abstract:The accuracy of pedestrian gait detection is critical for personal navigation system. In view of the problem that the conventional gait detection algorithm in the current pedestrian autonomous navigation system cannot be applied to the gait detection in the multiple motion states, an adaptive gait detection algorithm based on MIMU is proposed in this paper.The algorithm first uses the threeaxis modulus variance, uniaxial variance difference and waveform phase of the accelerometer to identify four different walking states, including forward, fast running, backward and lateral walking, and then sets the adaptive thresholds for different walking states, thus the adaptive gait detection in all kinds of motion state can be realized.The experimental verification on the algorithm has been carried out by fixing the independentlydeveloped MIMU on the lumbar spine. The results show that the gait detection accuracy of the forward walking and fast running is up to 99%, and the gait detection accuracy of backward and laterally walking is up to 93%, which proves thatthe adaptive gait detection algorithm is suitable for the personal navigation system.