Binaural Speech Separation Algorithm Based on Long and Short Time Memory Networks

Speaker separation in complex acoustic environment is one of challenging tasks in speech separation.In practice,speakers are very often unmoving or moving slowly in normal communication.In this case,the spatial features among the consecutive speech frames become highly correlated such that it is helpful for speaker separation by providing additional spatial information.To fully exploit this information,we design a separation system on Recurrent Neural Network(RNN)with long short-term memory(LSTM)which effectively learns the temporal dynamics of spatial features.In detail,a LSTM-based speaker separation algorithm is proposed to extract the spatial features in each time-frequency(TF)unit and form the corresponding feature vector.Then,we treat speaker separation as a supervised learning problem,where a modified ideal ratio mask(IRM)is defined as the training function during LSTM learning.Simulations show that the proposed system achieves attractive separation performance in noisy and reverberant environments.Specifically,during the untrained acoustic test with limited priors,e.g.,unmatched signal to noise ratio(SNR)and reverberation,the proposed LSTM based algorithm can still outperforms the existing DNN based method in the measures of PESQ and STOI.It indicates our method is more robust in untrained conditions.

The finite element analysis of cemented long and short stem prosthetic replacement in aged patients with comminuted intertrochanteric fracture

Objective To investigate the stress distribution of the femur after cemented prosthetic replacement in aged patients with comminuted intertrochanteric fracture and to analyze the difference of stress distribution between cemented long

Interest Points Analysis for Internet Forum Based on Long-Short Windows Similarity

For Internet forum Points of Interest(PoI),existing analysis methods are usually lack of usability analysis under different conditions and ignore the long-term variation,which lead to blindness in method selection.To address this problem,this paper proposed a PoI variation prediction framework based on similarity analysis between long and short windows.Based on the framework,this paper presented 5 PoI analysis algorithms which can be categorized into 2 types,i.e.,the traditional sequence analysis methods such as autoregressive integrated moving average model(ARIMA),support vector regressor(SVR),and the deep learning methods such as convolutional neural network(CNN),long-short term memory network(LSTM),Transformer(TRM).Specifically,this paper firstly divides observed data into long and short windows,and extracts key words as PoI of each window.Then,the PoI similarities between long and short windows are calculated for training and prediction.Finally,series of experiments is conducted based on real Internet forum datasets.The results show that,all the 5 algorithms could predict PoI variations well,which indicate effectiveness of the proposed framework.When the length of long window is small,traditional methods perform better,and SVR is the best.On the contrary,the deep learning methods show superiority,and LSTM performs best.The results could provide beneficial references for PoI variation analysis and prediction algorithms selection under different parameter configurations.

Numerical simulation of rockburst disaster and control strategy of constant resistance and large deformation anchor cable in Gaoloushan tunnel

The Gaoloushan Tunnel in Longnan City,Gansu Province,China,frequently experiences rockburst disasters due to high in-situ stress.Managing rockburst in deep-buried tunnels remains a challenging issue.This paper employs RFPA(Rock Failure Process Analysis)software to establish a calculation model of constant resistance and large deformation(CRLD)anchorages and analyzes the effects of different support methods and pre-stress levels on rockburst.We simulate the process of tunnel rockburst disasters and find that ordinary anchor support incurs rockburst on the right arch waist and arch top,forming a V-shaped explosion pit.CRLD anchor support has several advantages in rockburst control,such as more uniform stress distribution in the surrounding rock,a uniform distribution of plastic zones,less noticeable damage to the tunnel,and effective control of the arch top displacement.The effectiveness of the CRLD anchor support under varying pre-stress conditions shows that a higher prestress results in a smaller plastic zone of the surrounding rock and arch top displacement and a lower number of acoustic emission signals,which better explains the excavation compensation effect.Moreover,adding long anchorages in the deep surrounding rock area can better control rockburst and reduce surrounding rock deformation.Based on these findings,we propose a comprehensive control system that combines long and short anchorages and provides the optimal scheme based on calculations.Therefore,by using high-prestress CRLD anchor support and the combination of long and short anchorages at critical positions,we can enhance the integrity of the surrounding rock,effectively absorb the energy released by the surrounding rock deformation,and reduce the incidence of rockburst disasters.

Fusion of Spiral Convolution-LSTM for Intrusion Detection Modeling

Aiming at the problems of low accuracy and slow convergence speed of current intrusion detection models,SpiralConvolution is combined with Long Short-Term Memory Network to construct a new intrusion detection model.The dataset is first preprocessed using solo thermal encoding and normalization functions.Then the spiral convolution-Long Short-Term Memory Network model is constructed,which consists of spiral convolution,a two-layer long short-term memory network,and a classifier.It is shown through experiments that the model is characterized by high accuracy,small model computation,and fast convergence speed relative to previous deep learning models.The model uses a new neural network to achieve fast and accurate network traffic intrusion detection.The model in this paper achieves 0.9706 and 0.8432 accuracy rates on the NSL-KDD dataset and the UNSWNB-15 dataset under five classifications and ten classes,respectively.

Deep Learning Network for Energy Storage Scheduling in Power Market Environment Short-Term Load Forecasting Model

In the electricity market,fluctuations in real-time prices are unstable,and changes in short-term load are determined by many factors.By studying the timing of charging and discharging,as well as the economic benefits of energy storage in the process of participating in the power market,this paper takes energy storage scheduling as merely one factor affecting short-term power load,which affects short-term load time series along with time-of-use price,holidays,and temperature.A deep learning network is used to predict the short-term load,a convolutional neural network(CNN)is used to extract the features,and a long short-term memory(LSTM)network is used to learn the temporal characteristics of the load value,which can effectively improve prediction accuracy.Taking the load data of a certain region as an example,the CNN-LSTM prediction model is compared with the single LSTM prediction model.The experimental results show that the CNN-LSTM deep learning network with the participation of energy storage in dispatching can have high prediction accuracy for short-term power load forecasting.

Effect of ground motion duration on inelastic displacement ratio of SDOF systems

In this paper,the influence of ground motion duration on the inelastic displacement ratio,C_(1),of highly damped SDOF systems is studied.For this purpose,two sets of spectrally equivalent long and short duration ground motion records were used in an analysis to isolate the effects of ground motion duration on.The effect of duration was evaluated for observed values of C_(1) by considering six ductility levels,and different damping and post-yield stiffness ratios.A new predictive equation of C_(1) also was developed for long and short duration records.Results of non-linear regression analysis of the current study provide an expression with which to quantify the duration effect.Based on the average values of estimated C_(1) ratios for long duration records divided by C_(1) for a short duration set,it is concluded that the maximum difference between long and short duration records occurs when the damping ratio is 0.3 and the post-yield stiffness ratio is equal to zero.

A novel approach to study generalized coupled cubic Schrödinger-Korteweg-de Vries equations

The Kortewegde Vries(KdV)equation represents the propagation of long waves in dispersive media,whereas the cubic nonlinear Schrödinger(CNLS)equation depicts the dynamics of narrow-bandwidth wave packets consisting of short dispersive waves.A model that couples these two equations seems in-triguing for simulating the interaction of long and short waves,which is important in many domains of applied sciences and engineering,and such a system has been investigated in recent decades.This work uses a modified Sardar sub-equation procedure to secure the soliton-type solutions of the generalized cubic nonlinear Schrödinger-Korteweg-de Vries system of equations.For various selections of arbitrary parameters in these solutions,the dynamic properties of some acquired solutions are represented graph-ically and analyzed.In particular,the dynamics of the bright solitons,dark solitons,mixed bright-dark solitons,W-shaped solitons,M-shaped solitons,periodic waves,and other soliton-type solutions.Our re-sults demonstrated that the proposed technique is highly efficient and effective for the aforementioned problems,as well as other nonlinear problems that may arise in the fields of mathematical physics and engineering.

Finger millet porridges subjected to different processing conditions showed low glycemic index and variable efficacy on plasma antioxidant capacity of healthy adults

Finger millet porridges(FMP),rich in nutrient and non-nutrient compounds have been used in the traditional food cultures in Asia.The aims of the study were to determine the effect of different processing conditions of finger millet grains on glycemic response,phenolic content and the antioxidant activities of FMP and to determine the short term and long term efficacy of its consumption on plasma antioxidant levels of healthy adults.Twelve types of FMP were prepared combining different processing conditions.Phenolic content of porridges as well as antioxidant activities were determined.The glycemic index(GI)and glycemic load(GL)values of FMP were also evaluated.The long term efficacy of FMP consumption on plasma glucose(PG),total cholesterol(TC)levels and plasma antioxidant capacity(PAC)of 18 subjects were investigated using a 24 weeks randomized cross-over study.The short term efficacy of porridge consumption on AC was determined.PAC was measured by trolox equivalent antioxidant capacity(TEAC)and ferric ion reducing antioxidant power(FRAP).All FMP exhibited low GI values(<55)except the raw roasted flour which showed high and medium GI values for both particle sizes used.Parboiling of finger millet grains with 15 min steaming produced FMP with low glycemic response and possessed high PAC.Compared to baseline,PAC measured using FRAP and TEAC assays increased after 8 weeks consumption of porridge though significant changes were not observed for PG and TC levels.Furthermore,PAC was increased by 23 and 14%after 2 h of porridge consumption as measured by TEAC and FRAP,respectively.FMP consumption increased the plasma total antioxidant capacity of healthy adults.Further research on examining the potential of FMP on improving the antioxidant capacity in patients with diabetes is warranted.

A Regularized LSTM Method for Predicting Remaining Useful Life of Rolling Bearings

Rotating machinery is important to industrial production. Any failure of rotating machinery, especially the failure of rolling bearings, can lead to equipment shutdown and even more serious incidents. Therefore, accurate residual life prediction plays a crucial role in guaranteeing machine operation safety and reliability and reducing maintenance cost. In order to increase the forecasting precision of the remaining useful life(RUL) of the rolling bearing, an advanced approach combining elastic net with long short-time memory network(LSTM) is proposed, and the new approach is referred to as E-LSTM. The E-LSTM algorithm consists of an elastic mesh and LSTM, taking temporal-spatial correlation into consideration to forecast the RUL through the LSTM. To solve the over-fitting problem of the LSTM neural network during the training process, the elastic net based regularization term is introduced to the LSTM structure.In this way, the change of the output can be well characterized to express the bearing degradation mode. Experimental results from the real-world data demonstrate that the proposed E-LSTM method can obtain higher stability and relevant values that are useful for the RUL forecasting of bearing. Furthermore, these results also indicate that E-LSTM can achieve better performance.

A deep network prediction model for heavy metal cadmium in the rice supply chain

Cadmium and its compounds are currently known as Class I carcinogens,and excessive intake can cause severe health damage to humans.Rice has a strong absorption effect on cadmium,and rice products with excessive cadmium content have caused several significant public health contamination incidents.It is essential to predict the development trend of cadmium hazards in the rice supply chain so that countermeasures can be formulated to reduce the hazards.This paper proposes a deep prediction model for cadmium hazards in the rice supply chain based on the regularization method.Firstly,a long and short-term memory network is used to build the depth prediction model by using the regularization method,and the noise penalty term is added to reduce the model fitting to the noise and prevent the over-fitting caused by the noise.Finally,the optimization of the model hyperparameters was carried out using a Bayesian optimization approach to develop the prediction performance.Then,early warning system for prediction of cadmium hazards in the rice supply chain is built based on the deep prediction model proposed in this paper with SOA architecture,including data resource,business logic,and application service layers.The proposed model performs well on an actual data set of cadmium hazards in the rice supply chain and fits the data well.