A method of pedestrian moving direction recognition based on mobile phone inertial sensor

Abstract

The invention provides a pedestrian moving direction identification method based on the inertial sensor of a mobile phone, that is, four states of pedestrian forward, backward, left movement and right movement are identified. The core of the present invention is a deep neural network model, which consists of three convolutional layers, two LSTM units, an attention mechanism module and a fully connected layer. After learning and training, the model can effectively mine sensors The information implicit in the data, and accurately determine the direction of the pedestrian's movement. The method of the invention has the advantages of fast calculation speed, high recognition accuracy, and is not easily affected by the environment, and the required cost is also low.

Description

technical field [0001] The invention describes a deep neural network method for recognizing pedestrian moving directions (forward, backward, left shift, and right shift) based on mobile phone inertial sensors, and belongs to the field of human behavior recognition. Background technique [0002] With the development of science and technology, human behavior recognition technology has brought more and more benefits in scientific research, production economy and life services, and has been paid more and more attention by scientists and scholars. The current human behavior recognition technology is mainly divided into two types: based on video images and based on inertial sensors. Due to the low cost of sensors, small amount of data and easy calculation, the human behavior recognition method based on inertial sensors has a good application prospect. However, most of the existing recognition technologies based on inertial sensors use traditional machine learning methods. Due to ...

AI Technical Summary

Problems solved by technology

However, most of the existing identification technologies based on inertial sensors use traditional machine learning methods. Due to the limitations of the size and power consumption of wearable sensors, the collected data will have large data noise, which makes traditional methods unable to To effectively carry out data mining, it is necessary to manually extract the features in the sensor data sequence, and the information that these features can express is limited, and the final recognition effect is very low, and prior experience is often added to assist judgment

Images