Search

More recently, Human Activity Recognition (HAR) based on sensors has become a hot topic due to its wide application. Researchers significantly reduce the cost of feature extraction and improve the accuracy of recognition by introducing deep learning networks. However, human activity data are greatly affected by the inter-personal variability which brings the interclass similarity and the intraclass diversity. They not only increase the difficulty of the closed-set classification, but also affect the performance on the open-set problem. To solve the problem, we design a framework using a loss function of Euclidean distance and a high-dimensional embedding layer to enhance the ability of deep learning networks to mine discriminative features. Furthermore, we define two kinds of open-set problems in HAR: the pseudo open-set problem and the completely open-set problem. We propose a new clustering method based on the Euclidean distance for the pseudo open-set problem, which reduces the computation cost and improves the accuracy. For the completely open-set problem ignored by other researches, we introduce the rejection score to evaluate the distance score between samples and all known classes, and realize the completely open-set classification. We conduct experiments using four common deep learning networks on three public datasets: OPPORTUNITY, PAMAP2 and UniMiB-SHAR. The results show that the performances of the model modified by our method are much better than those of the original model.

In this paper, a novel multi-stream spatial digital predistortion (DPD) technique is proposed to model and linearize the fully-connected (FC) hybrid beamforming (HBF) transmitters. The proposed scheme solves the DPD implementation issue in the FC HBF array by estimating and linearizing the beam signals instead of the individual PAs. In FC HBF systems, significant intermodulation (IMD) between different transmit signals will be generated due to the analog beamforming and combining network upstream of the power amplifiers (PAs). The IMD beams will end up being radiated in different directions and some of them might fall in the linear beam directions. Therefore, multi-input DPD blocks using a practical multi-variable model are constructed for each RF chain to eliminate the complicated inner- and cross-channel IMDs of the beam signals. Simulations on a 4-stream 64-element FC HBF array and experimental tests on a 2-stream 4-element system are carried out to benchmark the proposed DPD technique against the conventional techniques. Better than 13 dB adjacent channel power ratio (ACPR) improvement and 12 dB normalized mean square error (NMSE) improvement have been achieved by the proposed DPD technique.

In this paper, a ring-slotted spiral local resonant phononic crystal structure is investigated. The energy band characteristics and transmission losses of the structure are calculated using the finite element method. The numerical results show that the structure can generate bandgap in the range of 23.54–102.87 Hz. Combined with the modal coupling theory, the vibrational modes of the ring-slotted spiral local resonant phononic crystal are analyzed to reveal the bandgap formation mechanism of the structure. The effects of the various dimensional parameters and material hardness on the bandgap of the structure are analyzed by varying them. The results show that by varying the filling radius of silicone rubber in the ring-slotted spiral local resonant phononic crystal, the structure can be made to have an onset frequency of 19.42 Hz and generate a bandgap in the range of 19.42–78.16 Hz. A new idea is provided for the application of phononic crystal in low-frequency environment for vibration and noise reduction.

In this paper, the acoustic streaming effects of Sonic Black Hole (SBH) are deeply studied by using the numerical and simulation method. By solving Navier–Stokes equations of compressible fluid, we analyze the flow characteristics of acoustic medium in SBH excited by sound wave, and further discuss the acoustic streaming effects and sound wave capture mechanism of SBH. The phenomenon of simultaneous reduction of medium and high-frequency sound reflection and transmission is analyzed. Under the excitation of medium- and high-frequency sound waves, the interaction between fluid medium flow and SBH structure leads to the uneven internal velocity distribution in the sound propagation direction, which can lead to the phenomenon of sound wave capture and deceleration. For low-frequency sound waves, the velocity distribution of sound medium is uniform, and there is almost no change in velocity gradient, resulting in the uniform distribution of sound pressure without sound absorption. These theoretical and numerical results are in good agreement with the experimental results in the literature, and also verify our results. Finally, by improving the structure, increasing the complexity of sound medium flow and the gradient change in the direction of sound propagation, the sound absorption and insulation ability of the SBH can be further improved. This study reveals the sound transmission mechanism of SBH, which can provide a new idea for the suppression of low-frequency sound waves in SBH.

High-frequency (HF) transmitter implemented with the multioctave power amplifier (PA) is faced with the problem of high-order harmonics owing to the nonlinearity of the PA. In this letter, a complexity-reduced harmonic-cancellation digital predistortion (CR-HC-DPD) is proposed and a narrow pulse sampling (NPS) method is adopted for model identification. Compared with the previous HC-DPD scheme, the proposed CR-HC-DPD approach greatly decreases the number of sampling points used for modeling and, in turn, reduces the computational complexity. Measurement results on a class-AB GaN PA with a 600-Hz two-tone signal demonstrated that the CR-HC-DPD scheme canceled all the harmonics for being below −60 dBc with less than 5% complexity compared with the HC-DPD scheme.

In this paper, a new digital predistortion (DPD) scheme is proposed to model and linearize the fully-connected hybrid beamforming (HBF) transmitters. In order to address the implementation issue of conventional DPD architecture in HBF array, the proposed DPD scheme allows to estimate and linearize the beam signals instead of the individual PAs. Different from the single-input single-output (SISO) DPD solutions in sub-array based HBF array, the proposed technique constructs a series of multi-input DPD blocks parallel for each RF chain to model and eliminate the complicated cross-channel intermodulation and inner-channel distortion specifically in fully-connected HBF array. Experimental results on a 2-stream 4-element fully-connected beamforming array show a 13 dB ACPR improvement of the proposed DPD than the conventional SISO DPD solutions. To the best of our knowledge, this is the first time that the DPD solutions for linearizing the FC HBF (b) array are investigated.

Due to the increase of signal bandwidth in the current fifth-generation (5G) communication, the strong memory effects of the power amplifier (PA) seriously deteriorate the linearity of the system, while traditional digital predistortion (DPD) algorithms do not work well in this scenario. In this paper, a cascaded Memory Polynomial-Neural Network (MP-NN) model is proposed to model the PAs with strong memory effects. By including prior information, the input of the neural network is changed from input signals over a period of time to the basis functions of input signals. In this way, the training process converges faster and achieves good modeling accuracy. Experimental results on a 3.2-3.8 GHz Doherty PA show that the proposed model outperforms the traditional MP model, the adjacent channel power ratio (ACPR) performance is improved from -36.7dBc to-48.8dBc

In this paper, we adopt both numerical and experimental approaches to investigate the turbulent flow in a cavity Acoustic Black Hole (ABH), and the low frequency sound energy dissipation mechanism under turbulence. Through the simulation by thermoviscous acoustic module, we deeply study the turbulent development process of the fluid acoustic medium in the cavity ABH, and how the turbulent evolution determines the acoustic performance of the ABH. This dissipation mechanism caused by turbulence in ABH enables cavity acoustic black hole to isolate more than 90% low frequency sound waves in an ultra-low and ultra-broad band, which can provide a good technical solution for the isolation of low frequency sound waves. Our experimental results not only prove our simulation and numerical results, but also prove an important theoretical result of turbulence theory, that is, in fully turbulence state, the relationship between turbulence energy spectral density E and wave number k (k = 1/L, L is the characteristic size L) is E ~ k−3.

In acoustic black holes (ABH) at low frequencies, catastrophe sound energy flux behaviour occurs spontaneously if the tip size of the ABHs reaches a critical value. This report demonstrates that this catastrophe behaviour can be quantified by analysing the relative phase transitions using Landau’s phase-transition theory. Two catastrophe behaviours corresponding to two stable phases (states) were observed using analytical calculations and a numerical analysis. The sound energy flux was proportional to 3 power laws and square relationships with the tip diameters, respectively. The 3 power law relationships were proved using low-frequency experimental observations. For engineering applications, ABH structures with small tip diameters can isolate more than 11 dB of low-frequency sound waves below 1600 Hz. We associated this cataclysmic phase transition with the geometric structure of the ABHs.