Speech emotion recognition is a vital and challenging task that the feature extraction plays a significant role in the SER performance. With the development of deep learning, we put our eyes on the structure of end-to-end and authenticate the algorithm that is extraordinary effective. In this paper, we introduce a novel architecture ADRNN (dilated CNN with residual block and BiLSTM based on the attention mechanism) to apply for the speech emotion recognition which can take advantage of the strengths of diverse networks and overcome the shortcomings of utilizing alone, and are evaluated in the popular IEMOCAP database and Berlin EMODB corpus. Dilated CNN can assist the model to acquire more receptive fields than using the pooling layer. Then, the skip connection can keep more historic info from the shallow layer and BiLSTM layer are adopted to learn long-term dependencies from the learned local features. And we utilize the attention mechanism to enhance further extraction of speech features. Furthermore, we improve the loss function to apply softmax together with the center loss that achieves better classification performance. As emotional dialogues are transformed of the spectrograms, we pick up the values of the 3-D Log-Mel spectrums from raw signals and put them into our proposed algorithm and obtain a notable performance to get the 74.96% unweighted accuracy in the speaker-dependent and the 69.32% unweighted accuracy in the speaker-independent experiment. It is better than the 64.74% from previous state-of-the-art methods in the spontaneous emotional speech of the IEMOCAP database. In addition, we propose the networks that achieve recognition accuracies of 90.78% and 85.39% on Berlin EMODB of speaker-dependent and speaker-independent experiment respectively, which are better than the accuracy of 88.30% and 82.82% obtained by previous work. For validating the robustness and generalization, we also make an experiment for cross-corpus between above databases and get the preferable 63.84% recognition accuracy in final. INDEX TERMS 3-D Log-Mel, dilated CNN, residual block, center loss, BiLSTM, attention mechanism.

Rubber layers with air-filled cavities or local resonance scatters can be used as anechoic coatings. A lot of researches have focused on the absorption mechanism of the anechoic coatings. As the anechoic coatings are bonded to the hull of submarine, the vibration of the hull should not be neglected when the analysis of the absorption characters is carried out. Therefore, it is more reasonable to treat the anechoic coating and the backing as a whole when the acoustic performance is analyzed. Considering the effects of the steel plate backing, the sound absorption performances on different models of anechoic coatings are investigated in this paper. The Finite Element Method is used to illustrate the vibrational behaviors of the anechoic coatings under the steel backings by which the displacement contours is obtained for analysis. The theoretical results show that an absorption peak is induced by the resonance of the steel slab and rubber layer. At the frequency of this absorption peak, the steel plate and the coating vibrates longitudinally like a mass-spring system in which the steel slab serves for mass and the coating layer is the spring. To illuminate the effects of the steel slab backing on the acoustic absorption, the thicknesses of the steel slab and the anechoic layer are discussed. Finally, an experiment is performed and the results show a good agreement with the theoretical analysis.

No abstract

In the indoor environment, due to weak receiver signals, environmental noise, multipath interference, and non-line-of-sight propagation, the traditional positioning algorithms based on received signal strength indication (RSSI) have many problems, such as inaccurate positioning results, great dependence on the signal propagation path loss model, and high time and labor costs. This paper studied the wireless indoor positioning algorithm based on neural network. A weighted median-Gaussian filtering method is proposed to preprocess RSSI and establish a location fingerprint database. An indoor positioning algorithm based on an improved fast clustering algorithm combined with a Levenberg-Marquardt (LM) algorithm is proposed. The improved clustering algorithm is used to design the network structure, initialize the number of radial basis function (RBF) neurons, find the local density peak as the cluster center to achieve rapid clustering of samples, and adjust the parameters of the kernel function of the hidden layer neurons. And the LM algorithm is used for numerical optimization. In order to verify the performance of the algorithm, positioning experiments are performed in the library. The error rate was reduced by 26.2% compared with the RBF network. The positioning results data confirm the effectiveness and applicability of the proposed algorithm.INDEX TERMS Neural networks, improved fast density clustering, LM, RSSI, Wi-Fi indoor positioning, wireless localization.

Purpose To improve the robustness of missile control system and reduce the error, a missile attitude adaptive control method based on active disturbance rejection control technology (ADRC) and BP neural network is innovatively proposed. Design/methodology/approach ADRC improves the performance of the missile control system by estimating and eliminating the total disturbance of the system. BP neural network adjusts the parameters of ADRC controller according to the state of the system to realize adaptive control. Based on the control system and missile dynamics model, the convergence analysis of the extended state observer and the stability analysis of the closed-loop system after embedding BP neural network are given. Findings The simulation results show that the adaptive control method can adjust the coefficient of error feedback rate according to the system input, output and error change rate, which accelerates the response speed of missile attitude angle and reduces the attitude angle error. Practical implications BP–ADRC further improves the robustness and environmental adaptability of the missile control system. The BP–ADRC control method proposed in this paper is proved feasible. Originality/value Different from the traditional ADRC, the BP–ADRC feedback signal proposed in this paper uses the output signal and its rate of the closed-loop system instead of the system state quantity estimated by extended state observer (ESO). This innovative method combined with BP neural network can make the system output meet the requirements when ESO has errors in the estimation of missile dynamics model.