With the rapid growth of researches toward computer vision and pattern recognition, methods that based on convolutional neural network (CNN) have shown unique advantages on handwritten characters recognition, also provided impressive results. This paper proposes a model based on CNN to deal with matters of handwritten Chinese character recognition. Different with conventional recognition system, in this model, input images are preprocessed by median filtering to smooth and reduce noise. For testing the stability and performance of the model, two testes are managed respectively. In integral test, experimental results show that the accuracy rate of recognition approach to 90.91% after 5000 times training, mean square error is decreased to 0.0079 at last. Meanwhile, this system also has a good performance at real-time test.

Benzo(a)pyrene, as the main polycyclic aromatic hydrocarbon pollutant in marine oil spill pollution, has negative effects on marine ecology and human health. A facile and sensitive method of rapid benzo(a)pyrene detection in seawater is essential for marine conservation. In this paper, a novel immunosensor is fabricated using a multi-walled carbon nanotubes-chitosan composite loaded with benzo(a)pyrene antibody. This immunosensor is based on a biosensing assay mechanism that uses multi-walled carbon nanotubes-chitosan composites as conductive mediators to enhance electron transfer kinetics. Then, potassium ferricyanide was used as an electrochemical probe to produce an electrochemical signal for the voltammetric behavior investigation of the immune response by differential pulse voltammetry. Under optimal experimental conditions, the peak current change was inversely proportional to the benzo(a)pyrene concentration in the range of 0.5 ng⋅ml−1 and 80 ng⋅ml−1 with a detection limit of 0.27 ng⋅ml−1. The immunosensor was successfully applied to assay BaP in seawater, and the recovery was between 96.6 and 100%, which exhibited a novel, sensitive and interference-resistant analytical method for real-time water environment monitoring. The results demonstrate that the proposed immunosensor has a great potential for application in the monitoring of seawater.

This research proposes a seawater desalination system driven by photovoltaic and solar thermal energy for remote regions such as islands and seaside villages where fresh water is not accessible. The performance of this system is demonstrated through experiments, and the main concerns are the output of the photovoltaic power generation system, power quantity, water yield, and the loads under different solar irradiance and temperature. In this system, a PLC is used as the controller to adjust the water pump by the collection and processing of sensor data. A load switching time system is designed to select different operating schemes under different environments in order to save energy. The control method of this system is developed to ensure that the photovoltaic power generation system does not undervoltage while maintaining the normal operation of the desalination system. An improved Perturbation and Observation (P&O) algorithm is also proposed as a new Maximum Point Power Tracking (MPPT) method to solve the problem of misjudgment and oscillation after tracking the maximum power point (MPP) in the traditional P&O algorithm. The simulation test in the MATLAB/Simulink environment shows that when external irradiance changes, the improved P&O algorithm can track the MPP faster than the traditional P&O algorithm, and the amplitude of oscillation on the MPP is smaller. The hardware experiments show that this system can operate stably and flexibly, and it is capable of producing 5.18 kWh of electric energy and 335.81 kg of freshwater per day. The maximum yield of the unit can reach 565.75 kg per day and the maximum daily power generation is 8.12 kWh.

This paper describes MixGAN-TTS, an efficient and stable non-autoregressive speech synthesis based on diffusion model. The MixGAN-TTS uses a linguistic encoder based on soft phoneme-level alignment and hard word-level alignment approach which explicitly extracts word-level semantic information, and introduces pitch and energy predictors to optimally predict the rhythmic information of the audio. Specifically, we use the GAN to replace the Gaussian function to model the denoising distribution, aiming to enlarge the denoising steps size and reduce the number of denoising steps to accelerate the sampling speed of diffusion model. Diffusion model using GAN can significantly reduce the denoising steps, and to some extent solve the problem of not being able to apply in real-time. The mel-spectrogram is converted into the final audio by the HiFi-GAN vocoder. Experimental results show that the MixGAN-TTS outperforms the other models compared in terms of audio quality and mel-spectrogram modeling capability for 4 denoising steps. The ablation studies demonstrate that the structure of MixGAN-TTS is effective.INDEX TERMS Speech synthesis, diffusion model, mixture attention mechanism, deep learning.