CN-122004830-A - Modularized human body posture recognition system based on plantar pressure and inertial sensor

Abstract

The invention belongs to the technical field of biological signal measurement and pattern recognition, in particular to a modularized human body posture recognition system based on plantar pressure and an inertial sensor, a system and a method for realizing high-robustness and high-interpretability human body whole body posture recognition through modularized layering decision, the invention aims to solve the limitation of the existing single sensor mode in human body posture sensing, and a sensing frame for multi-source heterogeneous information fusion is constructed by innovatively and cooperatively using a pressure sensor array deployed on the sole of a foot and an inertial measurement unit worn on the trunk.

Inventors

• BAI YU
• LV WENQI
• WANG YANG

Assignees

• 长春理工大学

Dates

Publication Date

20260512

Application Date

20260204

Claims (5)

1. 1. A modular human body posture recognition system based on plantar pressure and inertial sensors, comprising: The lower body gesture sensing module is used for identifying and outputting N basic lower body gestures based on plantar pressure distribution information; the upper body gesture sensing module is used for identifying and outputting M basic upper body gestures based on the trunk inertia measurement information; And the central processing module is pre-stored with a gesture combination mapping rule base and is used for receiving the basic lower body gesture and the basic upper body gesture, mapping the combination of the basic lower body gesture and the basic upper body gesture into a global human body gesture according to the rule base and outputting the global human body gesture.
2. 2. The modular human body posture recognition system based on plantar pressure and inertial sensors of claim 1, wherein the lower body posture sensing module comprises a flexible plantar pressure sensor array worn on the feet, and the basic lower body posture comprises a bipedal equilibrium support, a left foot dominant support, a right foot dominant support, and a bipedal ground clearance.
3. 3. The modular body position recognition system based on plantar pressure and inertial sensors of claim 1, wherein the upper body position sensing module comprises at least one Inertial Measurement Unit (IMU) worn on the torso of the person, and wherein the basic upper body position comprises torso upright, forward leaning, backward leaning, sideways leaning.
4. 4. The modular human body gesture recognition system based on plantar pressure and inertial sensors of claim 1, wherein the gesture combination mapping rule base is a predefined lookup table or rule set, which is input as a combination of upper body gesture tags and lower body gesture tags, and output as a global gesture tag with explicit semantics.
5. 5. The modular human body posture recognition system based on plantar pressure and inertial sensors of claim 1, wherein the central processing module further comprises a time sequence processing unit for smoothing or motion recognition of the global human body posture by a state machine or a time sequence model in combination with a basic posture sequence of a plurality of continuous moments.

Description

Modularized human body posture recognition system based on plantar pressure and inertial sensor Technical Field The invention belongs to the technical field of biological signal measurement and pattern recognition, and particularly relates to a modularized human body posture recognition system based on plantar pressure and an inertial sensor. Background Human body gesture recognition is used as a key enabling technology, has wide application prospects in various fields, such as being used for optimizing the action gesture of athletes and improving training efficiency in sports science, being used for gait analysis, balance ability assessment and postoperative recovery monitoring in the rehabilitation medical field, realizing natural and immersive control experience in Augmented Reality (AR), virtual Reality (VR) and intelligent man-machine interaction, and playing an important role in security monitoring, senile fall early warning and daily behavior analysis. The current mainstream technical route mainly depends on computer vision and various wearable sensors. Although the method based on computer vision can realize non-contact and large-scale gesture sensing, the performance of the method is extremely easy to be restricted by environmental illumination change, background interference, observation shielding and privacy protection requirements, and generally depends on high hardware and huge computing resources, so that the method is difficult to be widely applied to scenes with limited movement, privacy or resources. On the other hand, solutions based on wearable sensors, in particular Inertial Measurement Units (IMUs), are favored because of their small volume, low power consumption and ease of wear, however, the sensor drift problems inherent to IMUs can cause integration errors to accumulate over time, severely affecting the long-term accuracy of absolute pose estimation, especially in static or low-speed motion states, with limited ability to distinguish between basic poses such as "standing", "sitting", "lying" and the like. In recent years, the development of flexible electronic technology has promoted the application of plantar pressure sensors in gesture recognition, and the plantar pressure sensors can directly and accurately capture the mechanical interaction information between a human body and a supporting surface and reflect the balance state and gait phase, but the complicated motion gesture of the upper half parts of the body such as the trunk, the upper limbs and the like is difficult to completely represent only by pressure distribution. Although there have been studies attempting to multi-modal data fusion of pressure sensors with IMUs, existing methods have tended to simply splice raw data or shallow features, and then input complex "end-to-end" black box models such as deep learning networks for decision making. The method has extremely high requirements on the quantity and quality of training data and poor model interpretation, is difficult to flexibly adapt to individual differences of different users, newly-increased identification types or customization requirements of specific application scenes, and has challenges on the robustness and expansibility of the system. Therefore, a new model of human body whole body gesture recognition, which can deeply fuse multi-source heterogeneous sensing information, has a clear and transparent structure, and has high precision, strong robustness and good expandability, is needed in the industry. Disclosure of Invention Aiming at the problems that the existing fusion scheme depends on a complex black box model, has poor expandability and is insufficient in adaptability to individual differences, the invention provides a recognition system and a recognition method which are transparent in structure and easy to expand, and aims to realize human body gesture recognition with high accuracy, high robustness and fast adaptation to new scenes by fully playing the complementary advantages of a multi-source sensor. In order to achieve the purpose, the invention provides a technical scheme of modularized layered decision fusion, which has the core ideas of 'regional perception and logical combination'. The invention provides a modularized human body posture recognition system based on plantar pressure and an inertial sensor, which is characterized by comprising the following components: 1. The lower body posture sensing module is composed of a flexible plantar pressure sensor array worn on the feet and used for collecting plantar pressure distribution signals in real time and classifying the plantar pressure distribution signals into limited 4 basic lower body postures based on preset rules, namely a biped balanced support, a left foot leading support, a right foot leading support and biped ground separation. 2. The upper body posture sensing module is composed of at least one Inertial Measurement Unit (IMU) worn on the human trunk (such as the chest or wais