A Pet Motion Recognition Method for Embedded Platform

Abstract

The invention discloses a pet motion recognition method oriented to an embedded platform, which performs sliding window processing on the three-axis acceleration signals acquired by the three-axis accelerometer, and for each sliding window, extracts the most violent axis through the acceleration range, and Calculate the signal period of this axis, and use the extreme difference of the axis signal with the most violent fluctuation of each sample and the signal period of this axis as the characteristics of the sample; generate a decision table through the characteristics of the training sample and the motion category to which the training sample belongs, and finally use the decision table according to the characteristics of the test sample Determine the sport category to which the test sample belongs. The motion recognition process of the present invention does not involve floating-point numbers, multiplication and division operations at all, and is very suitable for use in embedded systems. The method of the present invention concentrates the signal analysis on the axis with the most violent fluctuations, so that the single-axis processing method can reduce the amount of computation and memory usage, and reduce the instability caused by the rotation of the three-axis accelerometer during the movement process, which is accurate and robust The advantages of high performance and real-time performance.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition and artificial intelligence, and in particular relates to a pet movement recognition method oriented to an embedded platform. Background technique [0002] With the popularization of mobile smart terminals such as mobile phones and tablets, mobile smart devices with acceleration sensors included in them have been widely used in people's lives. Mobile smart devices with acceleration sensors can easily identify human body Human motion recognition is a hot issue in the field of computer vision, and has broad application prospects and potential economic value in the fields of intelligent health, human-computer interaction, and video retrieval. At present, with the rise of human motion recognition, all kinds of smart watches and bracelets emerge in an endless stream, which can recognize people's motion status in real time, which also makes people start to pay more attention to their own hea...

AI Technical Summary

Problems solved by technology

At the same time, the embedded system also has many constraints on the pet motion recognition and analysis system, mainly including three aspects: storage space, computing power and power consumption: first, the memory space of the embedded system is very limited, if an algorithm needs to be integrated into the embedded The memory space occupied by the system must be strictly limited; secondly, the computing power of the embedded system, especially the computing power of high-precision floating-point numbers, is inefficient, and if there are a large number of floating-point number operations in the algorithm, it will inevitably increase Its overhead; finally, power consumption control of embedded systems is also a difficult point, excessive power consumption will seriously affect the battery life of the system

Images