Some recent studies use a convolutional neural network (CNN) or long short-term memory (LSTM) to extract gait features, but the methods based on the CNN and LSTM have a high loss rate of time-series and spatial information, respectively. Since gait has obvious time-series characteristics, while CNN only collects waveform characteristics, and only uses CNN for gait recognition, this leads to a certain lack of time-series characteristics. LSTM can collect time-series characteristics, but LSTM results in performance degradation when processing long sequences. However, using CNN can compress the length of feature vectors. In this paper, a sequential convolution LSTM network for gait recognition using multimodal wearable inertial sensors is proposed, which is called SConvLSTM. Based on 1D-CNN and a bidirectional LSTM network, the method can automatically extract features from the raw acceleration and gyroscope signals without a manual feature design. 1D-CNN is first used to extract the high-dimensional features of the inertial sensor signals. While retaining the time-series features of the data, the dimension of the features is expanded, and the length of the feature vectors is compressed. Then, the bidirectional LSTM network is used to extract the time-series features of the data. The proposed method uses fixed-length data frames as the input and does not require gait cycle detection, which avoids the impact of cycle detection errors on the recognition accuracy. We performed experiments on three public benchmark datasets: UCI-HAR, HuGaDB, and WISDM. The results show that SConvLSTM performs better than most of those reporting the best performance methods, at present, on the three datasets.

Pedestrian dead reckoning (PDR) using inertial and magnetic measurement unit (IMMU) is a current research hotspot that can be used to provide location-based services. The main challenges of the self-contained sensor-based PDR method are mainly sensor errors and orientation estimation errors. In this paper, a novel orientation estimation algorithm and gait phase detection algorithm with strong adaptability are proposed. The variance and magnitude of the angular rate are adopted to detect the gait. The experimental results show that the gait phase detection method can accurately distinguish the swing phase and the stance phase of the foot even under running conditions. By reasonable assumptions, the attitude angles calculated in the swing phase can be calibrated while the attitude angles are in the stance phase. The calibration of velocity not only uses the zero velocity updates method but also makes reasonable assumptions about the bias error of the accelerometer, so that the velocity calculated during the whole swing phase can be calibrated. Experimental results demonstrate that the proposed self-contained IMMU is capable of providing consistent beacon-free PDR in different scenarios, achieving less than 1.2% average distance error and average end-to-end position error.

The Pedestrian dead reckoning (PDR) positioning has small noise and high stability in the short term, but it cannot use for long term because of its cumulative error. The Received Signal Strength (RSS) fingerprint positioning has wide coverage and no accumulative error, but the amount of data leads to the long positioning time and discontinuous trajectory in the fingerprint database, which the stability of wireless signal is poor to result in positioning error. Based on above method’s issues, this paper proposes a fusion indoor positioning algorithm based on PDR and RSS fingerprint. This algorithm presents a new filtering method based on the comparison of two adjacent RSS values, which can be used to deal with the PDR positioning method to solve the noise point caused by the shaking because the tester handheld terminal is not stable enough in the test and be used for processing of RSS fingerprint positioning method to inhibit the effect of wireless signal instability in a certain extent, such as the jumping point noise. Moreover, a fusion algorithm based on adaptive parameters is proposed in this paper with differences from the fusion method proposed in the reference literature, which is adaptive parameter with different weights according to the accuracy of the two positioning methods in the testing. Xiaomi Mi Pad of the Android system verifies the proposed method and experimental results show that the proposed fusion positioning algorithm can effectively improve the positioning accuracy.

To solve the problem of time-consuming, laborious and low accuracy in magnetic map construction, a trajectory guided two-dimensional magnetic map construction method is proposed in this paper. The absolute displacement calculated by the inertial navigation system is mapped to the real trajectory to construct the magnetic map. The precision of the constructed magnetic map can reach 0.1141 m, and the mapping efficiency is greater than 1800 m2/h. Moreover, the cost of sensor is only $15.4. To deal with the instability problem that the traditional particle filter in PDR suffers from, this paper proposes an integrated particle filter algorithm. The algorithm uses two assumptions: 1. All noises obey Gaussian distribution; 2. The initial position and the results obtained from integrated particle filter are reliable enough. The particle distribution in the case of infinite particles is simulated, and similar particles are integrated. The algorithm breaks through the limit of particle number of particle filter and greatly reduces the randomness of particle filter algorithm. The real-world experiments show that the proposed positioning method can achieve an average positioning error of 0.32m.