Tài liệu Implement step and heading estimation using sensors on smartphones

Vietnam National University Ho Chi Minh City HCM University of Technology Faculty of Computer Science and Engineering ——————– * ——————— Graduate Thesis Implement Step and Heading Estimation Using Sensors on Smartphones Council : Computer Engineering Supervisors : Pham Hoang Anh, Ph.D Le Thanh Van, Ph.D Reviewer : Pham Quoc Cuong, Assoc. Prof. Ph.D Students : Tran Quang Dai (1552089) Nguyen Tien Loc (1552214) Ho Chi Minh City, 10/2021 TRƯỜNG ĐẠI HỌC BÁCH KHOA KHOA KH & KT MÁY TÍNH CỘNG HÒA XÃ HỘI CHỦ NGHĨA VIỆT NAM Độc lập - Tự do - Hạnh phúc ---------------------------Ngày 06 tháng 08 năm 2021 PHIẾU CHẤM BẢO VỆ LVTN (Dành cho người hướng dẫn/phản biện) 1. Họ và tên SV: Trần Quang Đại, Nguyễn Tiến Lộc MSSV: 1552089, 1552214 Ngành (chuyên ngành): Kỹ thuật Máy tính 2. Đề tài: Implement Step and Heading Estimation Using Sensors on Smartphones 3. Họ tên người hướng dẫn/phản biện: Phạm Hoàng Anh, Lê Thanh Vân 4. Tổng quát về bản thuyết minh: Số trang: Số chương: Số bảng số liệu Số hình vẽ: Số tài liệu tham khảo: Phần mềm tính toán: Hiện vật (sản phẩm) 5. Tổng quát về các bản vẽ: - Số bản vẽ: Bản A1: Bản A2: Khổ khác: - Số bản vẽ vẽ tay Số bản vẽ trên máy tính: 6. Những ưu điểm chính của LVTN: - Basically, the students have successfully studied and implemented an algorithm for step and heading estimation using only built-in sensors on smartphones, which is still a challenge to achieve high accuracy regarding this topic because of device dependency. The students have also performed various experiments to evaluate their implementation, and the accuracy errors are acceptable. - The students have demonstrated their capability in self-studying and then apply new knowledge and techniques to implement the proposed system. - The report has been well written. However, the students should revise the details. 7. Những thiếu sót chính của LVTN: - The students should investigate and implement various algorithms for having a comparative study, which helps to choose a suitable algorithm for a specific application. - The mobile application is just for testing algorithms since it is still simple without UI as an end-user application. 8. Đề nghị: Được bảo vệ o Bổ sung thêm để bảo vệ o 9. Câu hỏi SV phải trả lời trước Hội đồng: 10. Đánh giá chung (bằng chữ: giỏi, khá, TB): Không được bảo vệ o Điểm : Trần Quang Đại 8.5/10 Nguyễn Tiến Lộc 7.5/10 Ký tên (ghi rõ họ tên) Phạm Hoàng Anh TRƯỜNG ĐẠI HỌC BÁCH KHOA KHOA KH & KT MÁY TÍNH CỘNG HÒA XÃ HỘI CHỦ NGHĨA VIỆT NAM Độc lập - Tự do - Hạnh phúc ---------------------------Ngày 07 tháng 8 năm 2021 PHIẾU CHẤM BẢO VỆ LVTN (Dành cho người phản biện) Trần Quang Đại – 1552089 – Ngành Kỹ thuật Máy tính Nguyễn Tiến Lộc – 1552214 – Ngành Kỹ thuật Máy tính 2. Đề tài: Implementation of Step and Heading Estimation Using Sensors on Smartphones 3. Họ tên người hướng dẫn/phản biện: Phạm Quốc Cường 4. Tổng quát về bản thuyết minh: Số trang: Số chương: Số bảng số liệu Số hình vẽ: Số tài liệu tham khảo: Phần mềm tính toán: Hiện vật (sản phẩm) 5. Tổng quát về các bản vẽ: - Số bản vẽ: Bản A1: Bản A2: Khổ khác: - Số bản vẽ vẽ tay Số bản vẽ trên máy tính: 6. Những ưu điểm chính của LVTN: - Students proposed an approach to estimating the steps and heading of pedestrians. - Students implemented the proposal well and tested the proposed systems with different scenarios. - The report is adequately written, although some sections need more details explained. 1. Họ và tên SV: 7. Những thiếu sót chính của LVTN: - Students need to improve the algorithm so that users can keep their mobile phones in their pockets instead of holding them in their hands. - Students should do more tests and compare results with other devices (like a smartwatch) or mobile apps to prove the concept. Besides, students should conduct testing with different working conditions such as running or biking to verify the application's ability. - Teamwork is not defined well when the contributions of students are not balanced. 8. Đề nghị: Được bảo vệ o Bổ sung thêm để bảo vệ o Không được bảo vệ o 9. 3 câu hỏi SV phải trả lời trước Hội đồng: a. Would you please explain why users must hold their phones in their hands while walking to measure steps? b. How can the algorithm recognize if users are walking or biking at low velocity? 10. Đánh giá chung (bằng chữ: giỏi, khá, TB): - Trần Quang Đại - Nguyễn Tiến Lộc Điểm: 8.5/10 Điểm: 7.5/10 Ký tên (ghi rõ họ tên) Phạm Quốc Cường Declaration We declare that this thesis represents our own work, except where due reference is made. Any contribution made to the research by others, with whom we have worked at Bach Khoa university or elsewhere, is explicitly acknowledged in the thesis. We also declare that this thesis has not been previously included in a thesis or dissertation submitted to this or any other institution for a degree, diploma or other qualifications. Tran Quang Dai, Nguyen Tien Loc Acknowledgement In order to complete this graduate thesis, our group have received a great deal of support and assistance. First, we would like to express our gratitude to our supervisors, PhD. Pham Hoang Anh and PhD. Le Thanh Van. They are the main direct instructors, providing materials as well as monitoring the progress of the topic and providing support when our group is in trouble. They kindly supported us and continuous advice went through the process of completion of our thesis. We could not forget all the effort and dedication of lecturers in the Department of Computer Science and Engineering in particular, as well as the Ho Chi Minh City University of Technology in general. We would particularly thank our form teacher, PhD. Pham Quoc Cuong, who has supported our class for the past few years. The knowledge received from the teachers is very valuable and useful, greatly supporting us in completing this graduation thesis. Lastly, our deepest thanks come to our family, relatives, and friends during the time we struggled with this thesis. Their kind help, care, encouragement, and motivation gave us strength and lifted us up both physically and mentally. Tran Quang Dai, Nguyen Tien Loc i Abstract Positioning estimation systems play an important role in everyday life. The global positioning system (GPS) is one of the most popular positioning systems, which is highly efficient for outdoor environments. However, in indoor scenarios, GPS signal reception is weak. Therefore, achieving good position estimation accuracy is a challenge. To overcome this challenge, it is necessary to utilize position estimation systems for indoor localization. This thesis aims to implement a step and heading system (SHS) for estimating position in indoor environments using handheld smartphones. SHS is a combination of three components: step detection, step length estimation, and heading estimation. ii Contents Declaration Acknowledgement Abstract i ii List of Figures Terms vii viii 1 Introduction 1 1.1 Topic Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Thesis Objective and Scope . . . . . . . . . . . . . . . . . . . . . . 3 1.3 Thesis Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 Literature Review 6 3 Theoretical Basis 10 3.1 3.2 Inertial Measurement Units (IMUs) . . . . . . . . . . . . . . . . . . 10 3.1.1 Accelerometer . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.1.2 Gyroscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.1.3 Magnetometer . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Pedestrian Dead Reckoning . . . . . . . . . . . . . . . . . . . . . . 12 iii Contents 3.3 3.4 Coordinates systems . . . . . . . . . . . . . . . . . . . . . . . . . . 14 3.3.1 Smartphone/Device Coordinates System . . . . . . . . . . . 14 3.3.2 Ground/Earth Coordinates System . . . . . . . . . . . . . . 15 3.3.3 Coordinates Transformation . . . . . . . . . . . . . . . . . . 16 Step Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.4.1 Peak Detection . . . . . . . . . . . . . . . . . . . . . . . . . 18 3.4.2 Zero-crossing Based Step Detection . . . . . . . . . . . . . . 18 3.4.3 Peak-trough based step detection . . . . . . . . . . . . . . . 19 3.5 Step Length Estimation . . . . . . . . . . . . . . . . . . . . . . . . 19 3.6 Heading Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.7 3.6.1 Gyroscope based heading estimation . . . . . . . . . . . . . 22 3.6.2 Principal Component Analysis for heading estimation . . . . 22 Position Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4 Methodology 4.1 4.2 24 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 4.1.1 Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 4.1.2 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 4.1.3 Expo framework 4.1.4 Device Motion from Expo . . . . . . . . . . . . . . . . . . . 26 . . . . . . . . . . . . . . . . . . . . . . . . 25 Planning and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 26 4.2.1 Step detection . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.2.2 Step length estimation . . . . . . . . . . . . . . . . . . . . . 30 4.2.3 Heading estmation . . . . . . . . . . . . . . . . . . . . . . . 30 4.3 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4.4 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.4.1 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 iv Contents 4.4.2 DeviceMotion Implementation . . . . . . . . . . . . . . . . . 34 4.4.3 Coordinates transformation . . . . . . . . . . . . . . . . . . 35 4.4.4 Data filtering . . . . . . . . . . . . . . . . . . . . . . . . . . 38 4.4.5 Step and Heading Detection . . . . . . . . . . . . . . . . . . 41 4.4.6 Step Detection . . . . . . . . . . . . . . . . . . . . . . . . . 41 4.4.7 Step Length Estimation . . . . . . . . . . . . . . . . . . . . 42 4.4.8 Heading determination . . . . . . . . . . . . . . . . . . . . . 43 4.4.9 Position Estimation . . . . . . . . . . . . . . . . . . . . . . . 43 5 Evaluation 46 5.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 5.2 Test Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 5.2.1 Test Scenario 1 . . . . . . . . . . . . . . . . . . . . . . . . . 50 5.2.2 Test Scenario 2 . . . . . . . . . . . . . . . . . . . . . . . . . 51 5.2.3 Test Scenario 3 . . . . . . . . . . . . . . . . . . . . . . . . . 52 5.2.4 Test Scenario 4 . . . . . . . . . . . . . . . . . . . . . . . . . 53 5.2.5 Test Scenario 5 . . . . . . . . . . . . . . . . . . . . . . . . . 54 5.2.6 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 6 Conclusion and future work 57 6.1 The contribution of the thesis . . . . . . . . . . . . . . . . . . . . . 57 6.2 Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 6.3 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 v List of Figures 1.1 Results of localization with two classes of location sensing (a) Localization base on Wi-Fi access points (b) Localization using IMU devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 3.1 General Architecture of Pedestrian Dead Reckoning System . . . . 13 3.2 Smartphone coordinate system . . . . . . . . . . . . . . . . . . . . 14 3.3 The positive directions of 3 axes of the smartphone . . . . . . . . . 15 3.4 Ground/Earth Coordinate System . . . . . . . . . . . . . . . . . . 16 3.5 Basic Peak Detection Method . . . . . . . . . . . . . . . . . . . . . 18 3.6 Basic Zero-crossing Detection Method . . . . . . . . . . . . . . . . 19 4.1 Accelerations of 5 steps when holding smartphone parallel to the ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.2 Accelerations of 5 steps when holding smartphone perpendicular to the ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.3 Accelerations of 5 steps when holding smartphone arbitrary . . . . 28 4.4 Walking acceleration magnitude in hand and pocket . . . . . . . . 29 4.5 The general view of our system architecture . . . . . . . . . . . . . 32 4.6 Installation Expo CLI . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.7 Testing Expo CLI . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 4.8 Expo Go app in iOS App Store . . . . . . . . . . . . . . . . . . . . 34 4.9 Subscribe Device Motion Function . . . . . . . . . . . . . . . . . . 35 vi List of Figures 4.10 Device’s coordinates (red) and the Ground’s coordinates (blue) . . 36 4.11 Acceleration data when holding the phone perpendicular to the ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 4.12 Rotated acceleration data when holding the phone perpendicular to the ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 4.13 Acceleration data when holding the phone in arbitrary orientation 4.14 Rotated acceleration data when holding the phone in arbitrary orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 4.15 Moving Average Filter Function 4.16 acceleration data in z-axis before (green) and after (red) filtering . 40 4.17 Transfer Data Function . . . . . . . . . . . . . . . . . . . . . . . . 44 4.18 Testing Detect Movement Algorithm . . . . . . . . . . . . . . . . . 45 5.1 Calibrate Step Length Screen . . . . . . . . . . . . . . . . . . . . . 47 5.2 Main Screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 5.3 Export data to other applications or excel file . . . . . . . . . . . . 49 5.4 Exported Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 5.5 Result Scenario 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 5.6 Result Scenario 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 5.7 Result Scenario 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 5.8 Result Scenario 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 5.9 Result Scenario 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 38 . . . . . . . . . . . . . . . . . . . 40 vii Terms CLI . . . . . . . . . . . Command-Line Interface. DoF . . . . . . . . . . Degrees of Freedom. GPS . . . . . . . . . . Global Positioning System: a satellite navigation system used to determine the ground position of an object. IMU . . . . . . . . . . Inertial Measurement Unit. IoT . . . . . . . . . . . Internet of Things. IPS . . . . . . . . . . . Indoor Positioning System. LBS . . . . . . . . . . Location Based Service. LPS . . . . . . . . . . Local Positioning System. PCA . . . . . . . . . . Principal Component Analysis. PDR . . . . . . . . . Pedestrian Dead Reckoning. RFID . . . . . . . . . Radio Frequency Identification. SDK . . . . . . . . . . Software Development Kit. SHS . . . . . . . . . . Step and Heading System. UI . . . . . . . . . . . . User Interface. WLAN . . . . . . . Wireless Local Area Network. WPAN . . . . . . . Wireless Personal Area Network. ZUPT . . . . . . . . Zero Velocity Update. 1 Introduction 1.1 Topic Introduction Nowadays, a wide variety of positioning techniques have a significant impact on our modern life. The popularization of portable devices has brought this area new opportunities. Together with navigation, more and more smartphone applications in terms of Location-based Services (LBS) have achieved considerable success. Concerning outdoor navigation, GPS is the most accurate localization service for smartphones at present. However, the GPS-based positioning system has inherent limitations in its accuracy. A GPS can be used to detect the position of pedestrians or objects in outdoor environments, but in complex indoor environments, such as weak received GPS signals, severe multi-path effects, Non-Line-of-Sight conditions, high signal attenuation, etc, there are highly considerable challenges when using it. Normally, the accuracy of satellite or cellular network-based techniques is more than 10 meters, which is completely unacceptable. With such rough precision, tracking a target within a building and distinguishing among rooms is not in any way realistic [3]. Indoor positioning, on the other hand, has received more and more attention in recent years. Statistics show humans averagely spend over 70% of their time on indoor activities in modern life [3]. Along with the booming of various terminal devices, indoor positioning has become a really promising field, and the relevant applications based on smartphones, tablets, and other wearable objects are sprouting up in every aspect of everyday life. For example, an indoor positioning system for 1 1.1. Topic Introduction a museum or an exhibition hall can not merely display information according to visitors’ real-time location, but also show the way to find what they need, such as an ATM, WC, lift, or exit. When people arrive at a certain room or area, the projector and lighting or any other supporting facilities can be activated automatically. In the future, indoor positioning techniques could make physical distribution management more intelligent and would be an indispensable component of the Internet of Things (IoT). With the rapid development of smartphone technology and the growth in the number of smartphone users, the most popular method of handling indoor position tracking is by using the inertial and magnetic sensors embedded in smartphones, in a framework known as pedestrian dead reckoning (PDR) [1]. Smartphone based navigation systems have proven valuable in many applications and help overcome the limitations of satellite signals. The advantage of using a smartphone-based position estimation system is that it requires no additional peripheral devices except for the smartphone itself. The major applications for smartphone-based PDR systems are localization in indoor spaces such as shopping malls, airports, GPS-denied outdoor environments, etc. Another advantage of this positioning system is that it uses only embedded sensors in one system instead of integrating Wi-Fi signals, radio frequency signals, GPS, or vision cameras. The integration of other systems increases the system cost and complexity. Integration of Wi-Fi signals and vision cameras requires installation and maintenance of the infrastructure. The main disadvantage is the increasing errors of the system over time due to random noise and device errors, and a method able to correct them is required. In other words, the trajectory paths obtained using pedestrian dead reckoning have a problem with location error accumulation, and therefore can not hold their performance continuously. Pedestrian dead reckoning with Inertial Measurement Units (IMUs) is a method of estimating a person’s movement by detecting movements, estimating stride lengths, and determining motion directions using sensors in smartphones. In the field of location-based applications, estimating motion direction is a well-known difficult issue. The difficulty in implementing this PDR is that the IMUs are housed on a smartphone, and obtaining the pedestrian’s direction of motion is highly dependent on the smartphone’s orientation with respect to the walking direction. 2 1.2. Thesis Objective and Scope 1.2 Thesis Objective and Scope The identification of steps and direction of movement combined with other indoor navigation systems will help complete the indoor navigation application on mobile devices. Currently, there are Wi-Fi-based indoor positioning systems that take a long time to scan signals from Wi-Fi hot spots to provide high accuracy. This leads to not ensuring real-time responsiveness in real-time applications. For time response and error response for indoor positioning applications, estimation of steps and direction of movement using sensors available on mobile devices will be performed in time to scan signals from the Wi-Fi access point for error correction. Hybrid algorithms can combine the results in Fig. 1.1 to achieve more reliable navigation. These algorithms use localization to reduce the drift of IMU and use the IMU result to smooth the location estimates of the absolute location. Figure 1.1: Results of localization with two classes of location sensing (a) Localization base on Wi-Fi access points (b) Localization using IMU devices. The objective of this thesis is to investigate and implement one solution 3 1.3. Thesis Structure for estimating steps and directions using sensors on mobile devices in order to evaluate the performance of algorithms that analyse data obtained from sensors on mobile devices. It should utilize the sensors that are common on modern mobile devices and interfere as little as possible with user habits. However, in the most realistic example, the device’s heading is not the same as the user’s heading; since the IMUs are positioned within a phone, its orientation will change depending on the pedestrian’s hand movements. As a result, the heading direction indicated by the IMU (or the smartphone) can differ from the walking direction. The PDR method itself is not very accurate, but can often be used to extend navigation to areas where other navigation systems are not available or combined with other methods to improve navigation accuracy. The accuracy of the PDR method is limited by accurate step detection, accurate estimation of step length, and accurate estimation of heading angle. Another challenge is distinguishing walking from running and other activities (such as biking, climbing stairs, or riding elevators). Our system will use common embedded sensors. This system is limited to detecting position changes in the horizontal plane and the mobile device is restricted to handheld without swinging while walking or being put in a pocket. The reasons will be clearly explained in section 4.2. The performance of this scheme will show the benefits and drawbacks of using this form of position measurement with the SHS, as well as various aspects that should be considered in future work. 1.3 Thesis Structure Fundamentally, this thesis consists of 6 chapters. These chapters are organized as follows: In chapter 2, the related research of indoor positioning system will be briefly introduced. In chapter 3, basic theoretical knowledge is introduced, which discusses the core concepts of PDR, namely step detection, step length and heading angle estimation, development of an algorithm for the SHS system with the help of inertial sensors, and other knowledge regarding to this system. 4 1.3. Thesis Structure In chapter 4, the methodology of the chosen positioning estimation algorithm is discussed in detail. In chapter 5, the analysis of both the experiments and results is provided. In chapter 6, all of the thesis is summarized and a few future works are therefore discussed. 5 2 Literature Review Regarding the aforementioned GPS challenge scenarios, many indoor positioning technologies have been implemented in the past two decades. Vision-based systems are a well-known option in surveillance and could be used for position tracking because the positioning methods from video analysis are considered accurate and real-time. However, the images captured by cameras would easily be sheltered from people’s bodies or other objects. Besides, some vision-based approaches rely on the installation of beacons. Meanwhile, it has high requirements for light intensity and the distribution of cameras. The limitations also lie in the comparative expensive equipment and the sizable data amount of optical flow. Similar drawbacks are found in the ultrasound and infrared methods as well. The techniques of Radio Frequency Identification (RFID) have been attempted in some experiments, but few of them can stably achieve the required meter-level accuracy in the mentioned scenarios above. Researchers have been more likely to adopt Wireless Local Area Network (WLAN) based technologies recently. The radio strength fingerprint in a certain building is analyzed statistically. Through some mathematical statistics methods, the positioning accuracy could be less than 2 meters, which is quite satisfactory for indoor scenarios. But the complete Wi-Fi coverage in the target building, as well as the real-time interaction between the users’ devices and the hot spots, is necessary. If there is no Wi-Fi access point available, or some access points fail temporarily, the WLAN-based solutions will also be ineffective. Bluetooth or Zigbee-based solutions have proved feasible in some experiments. In certain scenarios, the accuracies are so satisfactory that these Wireless Personal Area Network (WPAN) techniques are regarded as the ideal 6 solutions for shopping malls or exhibition halls. Basically, high density equipped beacons are necessary, undoubtedly. Moreover, their positioning effects are also strongly associated with realistic environments. Even the number of pedestrians could greatly influence the performance. Especially in some extreme situations, such as rescue missions in damaged buildings after earthquakes or fires, or archeological and scientific expeditions in caves or underground where the navigation signals from satellites or cellular networks are usually unavailable and the extra infrastructures such as WLAN are insufficient, some alternative self-contained strategies which do not rely on external signals have to be developed to continue tracking people’s positions [3]. The PDR method is used by research [2] to improve outdoor pedestrian tracking with smartphones, which performs better than the smartphone’s built-in GPS module in terms of accuracy. In [9], they focus on step recognition, using the built-in smartphone accelerometer pedometer method, mainly divided into threshold setting, peak detection, correlation analysis, or spectral analysis. The peak detection-based method does not use threshold-based conditions to calculate the number of steps but determines steps by calculating the peak value of the accelerometer reading. Generally, these methods first extract local peaks in the acceleration amplitude and then apply temporary constraints on the detected peaks to reduce overcounting. After this operation, each remaining peak value is counted as a step. However, during the transition from the step mode or the smartphone posture, the results of the method based on peak detection may be inaccurate. In the article [6], the authors successfully developed a high-precision step count for various gaits using a low-cost Three Degrees of Freedom (3DoF) accelerometer. By proposing four characteristics: minimum peak distance, minimum peak protrusion, dynamic threshold, and vibration elimination, combined with periodic features and similarity, they successfully solved the overcounting, undercounting, and incorrect operation problems. In addition, the model from [11] has another advantage. Their model can promote the stability of the entire positioning system. Once a peak or valley is misidentified, it can cause a series of disturbances in the peak-valley pair, leading to nonsensical results in the peak-valley-based model. But the variance can be measured with certainty, and the influence of the frequency of errors that can 7