In this paper, a wavelet packet (WP)‐based acoustic feature extraction approach is proposed for automatic speech emotion recognition (SER). First, the issue of optimising the WP filter‐bank structure for giving classification task is presented as a tree pruning problem, and different tree‐pruning criteria are investigated. On this basis, a novel WP‐based feature is introduced for SER, namely discriminative band WP power coefficients. Finally, a SER system is built and extensive experiments are carried out. Experimental results show that the proposed feature considerably improves emotion recognition performance over conventional mel frequency cepstrum coefficient (MFCC) feature. The proposed feature extraction approach is promising since it can be easily extended to two‐dimensional (2D) facial expression analysis with 2D WP quadtree structures, and further a high‐quality audio–visual bimodal emotion recognition system is desirable.

Noise is an unneglectable problem in emotion recognition if we want to put it into practice. First, aiming at the problem of noise in speech, we design a new acoustic feature, Long time frame Analysis Weighted Wavelet Packet Cepstral Coefficient (LW-WPCC), for better robustness. To extract LW-WPCC feature, first the best wavelet packet basis is constructed. On the basis of this, a robust wavelet packet Cepstral Coefficient is extracted by combining short time frame analysis with long time frame analysis. After that, we introduce a sub-band spectral center-of-mass parameter with good robustness to additive noise and propose an extraction algorithm of LW-WPCC. Through experiments on speech emotion recognition of different SNR levels, it is shown that our proposed method shows better noise robustness and performance on speech emotion recognition. What’s more, as facial expressions will not be affected by noise, we do bio-modal emotion recognition based on audio-visual data to improve robustness by making a decision-level fusion. Experiments based on audio-visual data are conducted to evaluate efficiency of our method. Results show that bio-modal emotion recognition based on audio-visual data can improve robustness and achieve better performance by benefiting from different kinds of emotion data.

Efficient anomaly detection is at the forefront of ensuring optimized operations and system safety, especially in the field of engineering. It is deemed indispensable for the enhancement of work efficiency, maximization of resource utilization, and proactive addressal of potential risks. From this practical perspective, a new method for granular representation of time series data is proposed in this study and is validated through a collective anomaly detection task. The density peak clustering (DPC) algorithm is employed to evaluate clusters in two-dimensional (2-D) data, which are created from the original dataset and its variations. Using the principle of justifiable granularity, each cluster is depicted, allowing the time series to be converted into rectangular information granules. A distinctive similarity measure, designed to assess the similarity among subsequences in this format, is then introduced, with subsequences having reduced similarity being pinpointed as anomalies. Extensive testing on various datasets confirms that the effectiveness and applicability of the proposed method are evident in real-world scenarios. This approach not only enhances the understanding of time series patterns but is also recognized for its prominence in engineering applications where precise anomaly detection is demanded. Due to its ability to capture the inherent structure of time series and analyze the correlation among data, superior effectiveness is consistently exhibited by the proposed approach when compared to other existing methods.

Pre-seismic anomaly detection plays a crucial role in reducing economic losses and casualties caused by earthquakes. This paper proposes a novel four-step approach for pre-seismic anomaly detection. In the first step, a series of pre-seismic features are extracted by analyzing the earthquake catalog and geomagnetic signals. In the second step, the multi-view learning strategy is employed to obtain fusion features. In the third step, multiple seismic stations in one seismic zone are treated as a seismic station network, and a pre-seismic anomaly detection model is constructed based on the station network. In the final step, four evaluation indicators are introduced to comprehensively evaluate the detection results. Verification results show that the proposed method is effective and achieves better performance than other existing methods.

Different sensors may experience degeneracy in specific scenarios, which can lead to failures in multi-sensor fusion optimization. To address the challenges of localization and mapping under such degeneracy conditions, this paper proposes a LiDAR-Inertial-Visual SLAM system with integrated degeneracy awareness. In the frontend, the system utilizes the ESIKF for tracking and incorporates a sensor degeneracy detection method to ensure that degraded sensor data do not affect the state error update. In the backend, we replace the conventional point cloud map with an adaptive voxel map to accelerate the retrieval of planar point cloud, and merge LiDAR planar features within a sliding window via region growing algorithm. Furthermore, the point-to-plane distance formula is optimized via the Rayleigh quotient theorem, which reduces the dimensionality of the optimization variables and simplifies the optimization process. Finally, we evaluate the proposed system on several challenging datasets and compare it with various state-of-the-art sensor fusion systems. The experimental results demonstrate that the proposed system achieves high localization accuracy in the majority of challenging scenarios.