Design of Speech Denoising Algorithm Based on Wavelet Threshold Function and PSO

At present, there are many shortcomings in the discontinuity of wavelet threshold function and the constant threshold of different decomposition layers and the constant error it produced. The amplitude-frequency characteristics of wavelet filters are studied and analyzed by mathematical modeling. An improved wavelet threshold function with adjustable parameters is proposed. Particle swarm optimization (PSO) algorithm is used to find the optimal parameters of the improved threshold function in a background noise environment. The improved wavelet threshold function is combined with Bayesian threshold method to obtain the threshold based on Bayesian criterion, which makes the threshold adaptive in different layers and overcomes the shortcomings of fixed threshold. Finally, the speech signal with optimal wavelet coefficients is obtained after reconstruction. Compared with the traditional threshold function, Simulation results show that the improved threshold function achieves precise notch denoising, effectively retains the singularity and eigenvalues of the signal, and reduces the signal distortion.

Data Processing in the Pilot Training Process on the Integrated Aircraft Simulator

Flight safety is an integral part of air transportation. Flight accidents are highly unlikely to appear but most of them are caused by the human factor. The aircrew training system for abnormal operations relies on integrated aircraft simulator-based exercises. Crew needs to be trained not to degrade piloting technique quality when facing increased psychophysiological tension. Therefore, methods evaluating the characteristics of ergatic aircraft control systems, warning systems for deterioration due to failures in avionics systems, piloting technique quality, and abnormal operation algorithms are necessary. An analysis of the bank angle has revealed that there are hidden increased tension manifestations in the human operator expressed in the transition of the flight parameter variation from a stationary random process to deterministic fluctuations in the form of a sinusoid. The goal of the research is to increase the efficiency of pilots’ training using integrated aircraft simulators based on the design and implementation of statistical data processing algorithms. To achieve the goal of the research, two algorithms for detecting deterministic fluctuations based on the Neyman-Pearson criterion and the optimal Bayesian criterion are developed. The presented algorithms can be used in the integrated simulator software to automate the decision-making process on piloting quality.

Standardization of CPUE for walleye pollock in the Okhotsk Sea with inclusion of some environmental factors

Catch of walleye pollock by Russia is the highest in the northern Okhotsk Sea where on average 0.94 million metric tons were caught annually in the period between 1962 and 2017, or around 24 % of the total yield of Russian fishery. The total stock and spawning stock of pollock grow there since 2002, though the catch per unit effort (CPUE) has significantly decreased in the beginning of 2018 despite expected high levels of both total and spawning stocks. The sea surface temperature, ice cover and storms frequency were examined as possible reasons of low fishing efficiency in 2018. For this purpose, the generalized linear models (GLM) and generalized additive models (GAM) of catch dynamics are compared. GAM with addition of temperature and storms factors has the lowest Schwarz’s Bayesian criterion and the highest explained deviance (61.6 %). Efficiency of fishing gears has nonlinear relationship with the towing time. CPUE has hypersensitivity to the stock biomass presented as the power dependence (γ = 0.94, r = 0.923). Standardized CPUE is recommended for using in the final GAM for the pollock stock assessment in the northern Okhotsk Sea, hypersensitivity of CPUE should be estimated and corrected if necessary.

Methods for Optimizing the Objects Detection Range by a Multi-Channel Optical-Electronic System

The paper describes the possibilities of increasing the objects detection range by a multichannel optoelectronic system, which can be done in two ways: by choosing a statistical criterion for making a decision and by a rational design of the receiving optical system. When analyzing ways to increase the quality of processing the signals received by the system through heterogeneous channels, we found several ways to improve the quality of decisions made: at the algorithmic, functional or design levels. In this study, we chose the methods of channel integration at the algorithmic and constructive levels. Since the problem of multi-alternative object detection by several spectral channels is being solved, the Bayesian criterion for the search quality of the minimum average risk was chosen at the algorithmic level as the most appropriate. The search for the minimum average risk was carried out according to the characteristics of the object type. The choice of this criterion when evaluating signals from two spectral ranges provided a gain in the detection range of about 10 %. At the design level, the function of the ratio of irradiance on channels, e.g. passive infrared and laser radar channels, is nonlinear. To ensure the same value of the increase in the detection range for each channel, it is necessary to redistribute the threshold sensitivity of the channels. We opted for a method of equalizing the ranges by changing the transmission coefficients of the channels. The alignment of the light-energy characteristics of the channels at the design level leads to a gain in the detection range of scanned objects by 5--7 %

Gradient-Based Optimization for Bayesian Preference Elicitation

Effective techniques for eliciting user preferences have taken on added importance as recommender systems (RSs) become increasingly interactive and conversational. A common and conceptually appealing Bayesian criterion for selecting queries is expected value of information (EVOI). Unfortunately, it is computationally prohibitive to construct queries with maximum EVOI in RSs with large item spaces. We tackle this issue by introducing a continuous formulation of EVOI as a differentiable network that can be optimized using gradient methods available in modern machine learning computational frameworks (e.g., TensorFlow, PyTorch). We exploit this to develop a novel Monte Carlo method for EVOI optimization, which is much more scalable for large item spaces than methods requiring explicit enumeration of items. While we emphasize the use of this approach for pairwise (or k-wise) comparisons of items, we also demonstrate how our method can be adapted to queries involving subsets of item attributes or “partial items,” which are often more cognitively manageable for users. Experiments show that our gradient-based EVOI technique achieves state-of-the-art performance across several domains while scaling to large item spaces.