Accurate Density Estimation Research Articles

Comparative analysis of seasonal wind power using Weibull, Rayleigh and Champernowne distributions

Accurate statistical modeling of wind speed variability is crucial for assessing wind energy potential, particularly in regions with low wind speeds and significant calm hours. This study evaluates the Champernowne distribution as a novel model for wind speed analysis, comparing its performance with the two-parameter Weibull, three-parameter Weibull, and Rayleigh-Rice distributions. Wind speed data at 10 m hub height over three years (2021–2023) from Ben Guerir, Morocco, were analyzed using statistical metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Bias Error (MBE), Coefficient of Determination (R2), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). The Champernowne distribution outperformed the other models across all metrics, achieving the lowest RMSE (0.00036), MAE (0.00022), AIC (650.52), and BIC (689.46), and the highest R2 (0.99998). Its ability to capture calm hours and extreme wind speeds provided more accurate power density estimates, with errors averaging 23%, compared to 33% and 42% for the Weibull and Rayleigh-Rice distributions, respectively. Despite low mean wind speeds (2.7 m/s), Ben Guerir’s ground-based power density ranged from 18 to 54 W/m2. The results suggest that conventional large-scale wind energy projects are unsuitable for Ben Guerir. Instead, small Vertical-Axis Wind Turbines (VAWTs) or alternative strategies should be considered. The Champernowne distribution’s robustness makes it a valuable tool for wind energy assessments, especially in regions with intermittent wind patterns, providing a foundation for more accurate modeling and energy planning.

Estimating in vivo power deposition density in thermotherapies based on ultrasound thermal strain imaging.

In thermal therapies, accurate estimation of in-tissue power deposition density (PDD) is essential for predicting temperature distributions over time or regularizing temperature imaging. Based on our previous work on ultrasound thermometry, namely, multi-thread thermal strain imaging (MT-TSI), this work develops an in vivo PDD estimation method. Specifically, by combining the TSI model infinitesimal echo strain filter with the bio-heat transfer theory (the Pennes equation), a finite-difference time-domain model is established to allow online extraction of the PDD. An alternating-direction implicit method is adopted to ensure numerical stability and computational efficiency in implementing the model. Based on simulations, the accuracy and effectiveness of the model are examined by comparing a preset PDD distribution with the estimated one. Then, TSI results are obtained from ultrasound data acquired in in vivo experiments; with the PDD estimated from that, TSI distributions are then "predicted" using a validated numerical procedure. The two TSI results are compared to verify the self-consistency of the proposed method. A simplified and more efficient protocol for obtaining an "equivalent spherical PDD" is also discussed.

Mobile Surveillance System using Unmanned Aerial Vehicle for Aerial Imagery

Crowd counting plays a vital role in public safety, particularly during riot scenarios where understanding crowd dynamics is crucial for effective decision-making and risk mitigation. Accurate crowd estimation in such environments enables authorities to monitor the situation in real time, allocate resources efficiently, and prevent potential escalations. However, counting individuals in a riot scenario presents unique challenges due to the chaotic nature of the scene, varying crowd densities, and obstructions caused by movement and environmental factors. Traditional methods struggle to provide reliable results in these conditions, necessitating advanced solutions. This study explores the implementation of CSRNet (Congested Scene Recognition Network), a state-of-the-art deep learning model, to address crowd counting in challenging environments characterized as "images in the wild." CSRNet’s ability to leverage dilated convolutions allows it to effectively capture contextual information and handle high crowd densities without sacrificing spatial resolution. We evaluate the model’s performance on diverse datasets, including aerial imagery and real-world riot scenarios, focusing on its adaptability to dynamic, unstructured environments. The results demonstrate the potential of CSRNet to provide accurate crowd density estimates under adverse conditions, offering critical insights for public safety applications. By addressing the technical challenges of implementing CSRNet in these contexts, this study contributes to the advancement of deep learning-based crowd counting, emphasizing its significance in real-world scenarios such as riots and other high-stakes events. Future work aims to further enhance the model's robustness and applicability to diverse operational settings.

Raccoon density estimation from camera traps for raccoon rabies management

AbstractDensity estimation for unmarked animals is particularly challenging, yet density estimates are often necessary for effective wildlife management. Raccoons (Procyon lotor) are the primary terrestrial wildlife reservoir for Lyssavirus rabies within the United States. The raccoon rabies variant (RRVV) is actively managed at landscape scales using oral rabies vaccination (ORV) within the eastern United States. To effectively manage RRVV, it is important to know the density of raccoons to appropriately scale the density of ORV baits distributed on the landscape. We compared methods to estimate raccoon densities from camera‐trap data versus more intensive capture‐mark‐recapture (CMR) estimates across 2 land cover types (upland pine and bottomland hardwood) in the southeastern United States during 2019 and 2020. We evaluated the effect of alternative camera configurations and durations of camera trapping on density estimates and used an N‐mixture model to estimate raccoon densities, including covariates on abundance and detection. We further compared different methods of scaling camera‐based counts, with the maximum number of raccoons seen on any given image within a day best explaining density. Camera‐trap density estimates were moderately correlated with CMR estimates (r = 0.56). However, densities from camera‐trap data were more reliable when classifying category of density as an index used to inform management (83% correct when compared to CMR estimates), although the densities in our study fell into the 2 lowest density classes only. Using more cameras reduced bias and uncertainty around density estimates; however, if ≤6 camera traps were used at a site, a line transect approach proved less biased than a grid design. Camera trapping should be conducted for at least 3 weeks for more accurate estimates of raccoon population density in our study area (<5% bias). We show that camera‐trap data can be used to assign raccoon densities to management‐relevant density index bins, but more studies are needed to ensure reliability across a greater range of environmental conditions and raccoon densities.

Use of a Carbon Density Growth Model to Assess the Potential Carbon Sink Function of a Mongolian Pine Plantation in Heilongjiang Province, Northeast China

Accurate estimation of the potential increase in the carbon (C) sink function of forests is required for climate mitigation and C neutrality assessments. Also, accurate forest carbon density estimates are critical for understanding national- and global-level carbon cycling and storage and can inform climate change mitigation. This study established a stand C density growth model to further analyze the C sink potential of planted Mongolian pine (Pinus sylvestris var. mongolica) forests. Samples (390) from fixed plots of Mongolian pine were collected in Heilongjiang Province, Northeast China. The site index (SCI) and stand density index (SDI) were introduced to a constructed stand C density growth model, with an optimal model selected through model fitting. The effect of SDI on stand C density in different SCI grouping intervals was assessed. Total C sequestration of Mongolian pine was calculated using the established C density model. Sample plots with higher C density in each forest age stand were selected to establish a model of potential C sequestration for Mongolian pine, and the difference between this rate and the average was compared to obtain the potential increase in C sink capacity of the forest stand. Slightly different fitting accuracies among the different C density growth models were observed, with the Richards model showing the best performance, which improved through the introduction of the SCI and SDI. Stand C density was associated with an increasing trend in SCI, which within each SCI subgroup was related to the increasing SDI trend. The potential C sequestration rate of the stand was close to the average between years 5 and 13. The average C sequestration rate peaked at 3.86 Mg·ha−1·year−1 at year 13, whereas the potential C sequestration rate peaked at 4.42 Mg·ha−1·year−1 in year 15. A gap between the potential and average C sequestration rate existed between ages 13 and 45, indicating the possibility for an increased C sink function in this forest age range. The Richards growth model incorporating SCI and SDI provided a better reflection of the C density of the Mongolian pine plantation, and the established stand C sequestration rate model showed that the optimal increment in the plantation C sink function can be obtained between years 13 and 45. The results of this study can guide C sink management in the Mongolian pine plantation.

Estimating vegetation structure and aboveground carbon storage in Western Australia using GEDI LiDAR, Landsat and Sentinel data

Abstract Worsening climate change impacts are amplifying the need for accurate estimates of vegetation structure and aboveground biomass density (AGBD) to assess changes in biodiversity and carbon storage. In Australia, increasing wildfire frequency and interest in the role of forests in the carbon cycle necessitates biomass mapping across large geographic extents to monitor forest change. The availability of spaceborne Light Detection and Ranging (LiDAR) optimised for vegetation structure mapping through the Global Ecosystem Dynamics Investigation (GEDI) provides an opportunity for large-scale forest AGBD estimates of higher accuracy. This study assessed the use of the GEDI canopy height product to predict woody AGBD across five vegetation types in Western Australia: tall eucalypt forests, eucalypt open‒woodlands, low-lying heathland, tropical eucalypt savannas, and tussock and hummock grasslands. Canopy height models were developed using random forest regressions trained on GEDI canopy height discrete point data. Predictor variables included spectral bands and vegetation indices derived from synthetic aperture radar (SAR) Sentinel‒1 data, and multispectral Landsat and Sentinel‒2 data. AGBD was subsequently estimated using power-law models derived by relating the predicted canopy heights to field AGBD plots. Mapping was conducted for 2020 and 2021. The accuracy of canopy height predictions varied with height quantiles; models underestimated the height of taller trees and overestimated the height of smaller trees. A similar underestimation and overestimation trend was observed for the AGBD estimates. The mean carbon stock was estimated at 69.0 ±12.0 MgCha-1 in the tall eucalypt forests of the Warren region; 33.8 ± 5.0 MgCha-1 for the open eucalypt woodlands in the South Jarrah region; 7.1 ± 1.4 MgCha-1 for the heathland and shrublands in the Geraldton Sandplains region; 43.9 ± 4.9 MgCha-1 for the Kimberley eucalypt savanna; and 3.9 ± 1.0 MgCha-1 for the Kimberley savanna grasslands. This approach provides a useful framework for the future development of this process for fire management, and habitat health monitoring.

Robust Estimation of Crowd Density Using Vision Transformers

Estimating and predicting crowd density at large events or during disasters is of paramount importance for enhancing emergency management systems. This includes planning effective evacuation routes, optimizing rescue operations, and ensuring efficient deployment of emergency services. Traditionally, surveillance systems that rely on cameras have been employed to monitor crowd movements. However, accurately estimating crowd density using such systems presents several challenges. These challenges stem primarily from the interaction between large crowds and the limitations of two-dimensional cameras in capturing the full scope of three-dimensional spaces. Optical distortions, environmental factors, and variations in camera angles further complicate the task, making accurate estimations difficult to achieve. To address these challenges, this paper introduces a robust method for calculating crowd density that leverages advanced vision transformers. By combining the output of these transformers with a two-stage neural network, the method effectively mitigates the limitations of traditional approaches. One of the key advantages of the proposed system is its robustness, which allows it to perform well across different camera specifications, installation locations, and image aspect ratios. The method applies and evaluates various deep learning techniques, introducing improvements to existing network structures that are better suited for the problem at hand. Extensive experimental verification demonstrates that the proposed method consistently produces accurate crowd density estimates, even in diverse and complex crowd environments. This robust performance underscores its potential for improving emergency management and crowd control in real-world situations.

Distortion corrected kernel density estimator on Riemannian manifolds

Manifold learning obtains a low-dimensional representation of an underlying Riemannian manifold supporting high-dimensional data. Kernel density estimates of the low-dimensional embedding with a fixed bandwidth fail to account for the way manifold learning algorithms distort the geometry of the Riemannian manifold. We propose a novel distortion-corrected kernel density estimator (DC-KDE) for any manifold learning embedding, with a bandwidth that depends on the estimated Riemannian metric at each data point. Exploiting the geometric information of the manifold leads to more accurate density estimation, which subsequently could be used for anomaly detection. To compare our proposed estimator with a fixed-bandwidth kernel density estimator, we run two simulations including one with data lying in a 100 dimensional ambient space. We demonstrate that the proposed DC-KDE improves the density estimates as long as the manifold learning embedding is of sufficient quality, and has higher rank correlations with the true manifold density. Further simulation results are provided via a supplementary R shiny app. The proposed method is applied to density estimation in statistical manifolds of electricity usage with the Irish smart meter data.

Evaluating sampling techniques for quantifying Asiatic garden beetle (Maladera formosae) (Coleoptera: Scarabaeidae) infestations in commercial mint.

The Asiatic garden beetle, Maladera formosae Brenske (AGB), has become a significant pest of commercial mint fields in northern Indiana. Larval feeding on mint roots can cause stunted growth and plant death when densities are high. Sampling approaches that provide reliable estimates of larval densities in mint have not been established, leaving farmers without the knowledge necessary to implement integrated pest management (IPM) strategies. To address this knowledge gap, we evaluated strategies for estimating AGB larval densities and plant performance in commercial mint systems. We used 2 sampling methods to collect larval density and plant performance data from 3 mint fields and conducted simulations to optimize sampling intensity (accuracy and precision) and sampling scheme (random vs. systematic) using these data. Additionally, we examined the sensitivity and efficiency of each sampling method. Compared to the cup-cutter method, the quadrat method provided the most accurate and precise estimates of larval density and plant performance, with ≤ 7 samples required per 0.2 ha. Quadrat excavation was also more sensitive, increasing the probability of detecting AGB larvae within a 32 m2 plot by 76.7%, and requiring significantly less time to survey an equivalent volume of soil for AGB larvae. When the quadrat method was employed, random sampling schemes provided below-ground biomass estimates that were significantly closer to the true mean of the sampling area. The results of this research will facilitate the development of IPM decision-making tools for farmers and support future research for AGB and other soil insect pests affecting mint production.

Computing the QRPA level density with the finite amplitude method

We describe a new algorithm to calculate the vibrational nuclear level density of an atomic nucleus. Fictitious perturbation operators that probe the response of the system are generated by drawing their matrix elements from some probability distribution function. We use the Finite Amplitude Method to explicitly compute the response for each such sample. With the help of the Kernel Polynomial Method, we build an estimator of the vibrational level density and provide the upper bound of the relative error in the limit of infinitely many random samples. The new algorithm can give accurate estimates of the vibrational level density. Since it is based on drawing multiple samples of perturbation operators, its computational implementation is naturally parallel and scales like the number of available processing units.

Floating-point histograms for exploratory analysis of large scale real-world data sets

Histograms are among the most popular methods used in exploratory analysis to summarize univariate distributions. In particular, irregular histograms are good non-parametric density estimators that require very few parameters: the number of bins with their lengths and frequencies. Although many approaches have been proposed in the literature to infer these parameters, most existing histogram methods are difficult to exploit for exploratory analysis in the case of real-world data sets, with scalability issues, truncated data, outliers or heavy-tailed distributions. In this paper, we focus on the G-Enum histogram method, which exploits the Minimum Description Length (MDL) principle to build histograms without any user parameter. We then propose to extend this method by exploiting a new modeling space based on floating-point representation, with the objective of building histograms resistant to outliers or heavy-tailed distributions. We also suggest several heuristics and a methodology suitable for the exploratory analysis of large scale real-world data sets, whose underlying patterns are difficult to recover for digitization reasons. Extensive experiments show the benefits of the approach, evaluated with a dual objective: the accuracy of density estimation in the case of outliers or heavy-tailed distributions, and the effectiveness of the approach for exploratory data analysis.

An improved solution of the probability density evolution equation for analyzing stochastic structural responses with adaptive time steps and initial conditions

The probability density evolution method (PDEM) offers an innovative approach for analyzing the stochastic vibration responses of structures. Given the complexities inherent in real-world challenges, the probability density evolution equations (PDEE) often require numerical methods for effective resolution. Consequently, improving computational efficiency and solution accuracy is of paramount importance. Currently, the impulse function discretization method (IFDM) faces several challenges, including limited differentiability in the spatial direction for initial value conditions. Furthermore, the probability density function produced by this method frequently deviates significantly from the true value, undermining the accuracy of the PDEE. Additionally, when addressing vibration responses characterized by rapid evolutionary velocities, the use of a globally fixed uniform time step size (FUTSS) results in a marked decrease in computational efficiency. To tackle these challenges, this study introduces an improved probability density evolution method (IPDEM), which integrates the Gaussian kernel density estimation discretization method (GKDEDM) and an adaptive non-uniform time step size (ANUTSS). The advantages of the IPDEM are illustrated through two numerical simulation examples. The findings indicate that the GKDEDM not only enhances spatial differentiability but also provides a more accurate estimation of the probability density at the initial moment, thereby improving the accuracy of the PDEE. Furthermore, the ANUTSS significantly reduces the number of iterations required while preserving precision, leading to a substantial optimization of computation time.

Rapid estimation of wood density and total extractives of two important Dalbergia species using near-infrared (NIR) spectroscopy and multivariate analysis

ABSTRACT Wood density and chemical extractives are important parameters influencing timber quality and utility in wood industry. Several wood properties, particularly mechanical parameters and durability, depend respectively on density and extractive contents. This study aimed to apply the near-infrared (NIR) spectroscopy technique for rapid and accurate estimation of density and extractive contents of Dalbergia latifolia and D. sissoo wood species. The objective was to develop a predictive model for properties, addressing the critical gap in efficient and large-scale assessment techniques for wood industries. Density and extractive contents were determined by conventional and wet chemistry methods. Spectral data of woods were subjected to multivariate analysis to improve the accuracy of calibration models and performance was assessed using coefficient of determination (R²) and root mean square error of prediction. The optimised models exhibited good predictive accuracy with high R²of 0.85 and 0.82 for density and extractive contents respectively. Spectral patterns correlating density and extractive variations identified provided valuable insights into the chemical composition and density relationship in Dalbergia species. The study thus highlights the potential of this non-invasive analytical approach in assessing the quality parameters, which would facilitate sustainable utilisation and contribute to the advancement of woodworking industries.

Settling velocity and effective density analysis for aquaculture floc particles: An approach through bivariate parametric copula

The study of flocs' characteristics in aquaculture tanks remains challenging due to their complex composition. Nevertheless, the application of experimental methods, such as particle tracking velocimetry (PTV), has made it possible to measure particle diameters and settling velocities. With the experimental data, bivariate distribution functions through copula modelling were employed to provide a more accurate estimation of the effective density of flocs through the empirical model proposed by Lau and Krishnappan. It was observed that the primary particle density constituting the flocs varied between 1051 kg/m³ and 1426 kg/m³. Moreover, when considering a variable primary particle density, effective floc density values ranging from 980 kg/m³ to 1500 kg/m³ were obtained for floc diameters ranging from 0.068 mm to 1.9 mm, respectively. This variation confirms that the b and c parameters of the Lau-Krishnappan model change for each floc diameter range following the quantiles associated with the conditional probability of diameter and settling velocity. Thus, employing copula approximation, a more accurate fit of the Lau-Krishnappan model was achieved, considering a wide range of particle diameters at the tails. This approach offers better estimates of floc effective density and settling velocity, essential for enhancing the selection and design of aquaculture tanks and settlers to ensure efficient solids removal.

Time-resolved measurements of the densities of individual frozen hydrometeors and fresh snowfall

Abstract. It is a challenge to obtain accurate measurements of the microphysical properties of delicate, structurally complex, frozen, and semi-frozen hydrometeors. We present a new technique for the real-time measurement of the density of freshly fallen individual snowflakes. A new thermal-imaging instrument, the Differential Emissivity Imaging Disdrometer (DEID), has been shown through laboratory and field experiments to be capable of providing accurate estimates of individual snowflake and bulk snow hydrometeor density (which can be interpreted as the snow-to-liquid ratio or SLR). The method exploits the rate of heat transfer during the melting of a hydrometeor on a heated metal plate, which is a function of the temperature difference between the hotplate surface and the top of the hydrometeor. The product of the melting speed and melting time yields an effective particle thickness normal to the hotplate surface, which can then be used in combination with the particle mass and area on the plate to determine a particle density. Uncertainties in estimates of particle density are approximately 4 % based on calibrations with laboratory-produced particles made from water and frozen solutions of salt and water and field comparisons with both high-resolution imagery of falling snow and traditional snowpack density measurements obtained at 12 h intervals. For 17 storms, individual particle densities vary from 19 to 495 kg m−3, and storm mean snow densities vary from 40 to 100 kg m−3. We observe probability distribution functions for hydrometeor density that are nearly Gaussian with kurtosis of ≈ 3 and skewness of ≈ 0.01.

Comparing camera‐based ungulate density estimates: a case study using island populations of bighorn sheep and mule deer

AbstractCamera‐based abundance estimators are an alternative methodology of growing interest in both research and management applications. The statistical formulations of camera‐based abundance estimators using time‐lapse data should theoretically produce precise and unbiased estimates; however, production of unbiased results also requires meeting several important assumptions, and real‐world case studies evaluating such results remain relatively few. We applied instantaneous sampling (IS) and space‐to‐event (STE) estimators to remote camera data collected in April 2021 via time‐lapse sampling of closed populations of bighorn sheep (Ovis canadensis) and mule deer (Odocoileus hemionus) on Wild Horse Island in western Montana, USA, and compared results for bighorn sheep to aerial and ground‐based counts. Point estimates from camera‐based approaches underestimated bighorn sheep populations by 32–44% (IS estimator) and 62–69% (STE estimator) relative to aerial and ground counts. Patchy spatial distribution and group‐living behavior of sheep resulted in a high degree of noise surrounding the IS estimate. In comparison, a low point estimate with relatively narrow confidence intervals suggested potential sensitivity of the STE estimator to violating assumptions of independence among individual animals and sampling occasions. Estimates of mule deer had improved precision over sheep estimates, as indicated by lower estimated coefficients of variation of the mean (CVmean) derived from the analytic SE estimator. Using 15‐m viewsheds and the IS estimators, mule deer density estimates came with a 26% CVmean compared to 43% CVmean for bighorn sheep. This discrepancy may be a result of differences in distribution, behavior, and relative abundance between the 2 species. Accounting for group size and increasing time between sampling may improve accuracy of density estimates and adhere better to model assumptions when estimating precision. In addition, factors influencing viewshed and resulting density extrapolations must be considered carefully. While camera‐based methods theoretically provide an alternative way to estimate density when traditional methods are impractical, our results suggest that more work is needed to ensure density estimates are accurate and precise enough to inform population management.

GalaxyFlow: upsampling hydrodynamical simulations for realistic mock stellar catalogues

ABSTRACT Cosmological N-body simulations of galaxies operate at the level of ‘star particles’ with a mass resolution on the scale of thousands of solar masses. Turning these simulations into stellar mock catalogues requires ‘upsampling’ the star particles into individual stars following the same phase-space density. In this paper, we introduce two new upsampling methods. First, we describe GalaxyFlow, a sophisticated upsampling method that utilizes normalizing flows to both estimate the stellar phase-space density and sample from it. Secondly, we improve on existing upsamplers based on adaptive kernel density estimation (KDE), using maximum likelihood estimation to fine-tune the bandwidth for such algorithms in a way that improves both the density estimation accuracy and upsampling results. We demonstrate our upsampling techniques on a neighbourhood of the Solar location in two simulated galaxies: Auriga 6 and h277. Both yield smooth stellar distributions that closely resemble the stellar densities seen in the Gaia DR3 catalogue. Furthermore, we introduce a novel multimodel classifier test to compare the accuracy of different upsampling methods quantitatively. This test confirms that GalaxyFlow more accurately estimates the density of the underlying star particles than methods based on KDE, at the cost of being more computationally intensive.

Assessment of alternative fluid calibration to estimate traceable liquefied hydrogen flow measurement uncertainty

The usage of liquefied hydrogen (LH2) is growing fast as the EU aims to be climate-neutral by 2050. Therefore, reliable and consistent measurements of LH2 amounts (as determined by flow meters) will become increasingly important for custody transfers. Reliability and consistency of measurements is established from calibration. Due to the lack of facilities being able to calibrate flow meters directly with LH2 as calibration fluid, the question arises whether the measurement error of the calibration performed on a different fluid and different process conditions are representative for the errors when measuring LH2. An on-site calibration of two turbine meters measuring LH2 flow was performed. A Coriolis mass flow meter calibrated on water was used as a reference in conjunction with a dedicated flow meter model, which was validated by a directly SI-traceable calibration on liquefied natural gas. LH2 flow rates were within 1000 kg/h and 3000 kg/h, which are relevant to hydrogen transport by LH2 trailers. The analysis of the measurement data reveal errors on totalized mass ranging from 2.0 % to 2.9 % with a total calibration uncertainty of 1.3 % (k=2), where the main uncertainty source was identified as the on-site density estimate necessary to convert the volume flow measurements into mass flow measurements. The results indicate the need for reliable and accurate temperature measurements (the latter allowing for an accurate density estimate), so that volume based and mass based amount measurements can be compared with smaller uncertainties.

Postfilter for Dual Channel Speech Enhancement Using Coherence and Statistical Model-Based Noise Estimation.

A multichannel speech enhancement system usually consists of spatial filters such as adaptive beamformers followed by postfilters, which suppress remaining noise. Accurate estimation of the power spectral density (PSD) of the residual noise is crucial for successful noise reduction in the postfilters. In this paper, we propose a postfilter utilizing proposed a posteriori speech presence probability (SPP) and noise PSD estimators, which are based on both the coherence and the statistical models. We model the coherence-based a posteriori SPP as a simple function of the magnitude of coherence between two microphone signals and combine it with a single-channel SPP based on statistical models. The coherence-based estimator for the PSD of the noise remaining in the beamformer output in the presence of speech is derived using the pseudo-coherence considering the effect of the beamformers, which is used to construct the coherence-based noise PSD estimator. Then, the final noise PSD estimator is obtained by combining the coherence-based and statistical model-based noise PSD estimators with the proposed SPP. The spectral gain function is also modified, incorporating the proposed SPP. Experimental results demonstrate that the proposed method led to more accurate noise PSD estimation and perceptual evaluation of speech quality scores in various diffuse noise environments, and did not degrade the speech quality under the presence of directional interference, although the proposed method utilizes the coherence information.

Electron density estimation using MPPC-based photon-counting CT for radiation therapy planning

Photon-counting computed tomography (PCCT) has been attracting attention recently, as a potential alternative to the conventional computed tomography (CT). With PCCT, the photon energy can be obtained, and the amount of exposure can be reduced compared with the conventional CT. We developed a detector system composed of a scintillator and multi-pixel photon counter (MPPC). This makes the system simple and inexpensive compared with the semiconductor-based PCCT, like CdTe and CZT. In this study, we propose estimating electron density for radiation therapy planning using a six-threshold MPPC-based photon-counting detector. Electron density is an essential information for determining beam energy in radiation therapy planning. Therefore, accurate electron density estimation is required. The estimation method used in this study is based on the method proposed by Saito et al. (2012) PCCT images of electron-density-known phantoms were acquired, and information of the two out of the six energy bands was used for estimation. As a result, we succeeded in estimating the electron density with an accuracy of 4.59% root-mean-square error. However, the lung phantoms, which had a relatively low electron density, still had a large deviation from the true value. In addition, dual-energy CT and PCCT estimation results were compared. PCCT results has not yet surpassed the accuracy of dual-energy CT (DECT) results; however, the high electron density materials were obtained almost at the same level of accuracy. In future, we will continue to work on more accurate electron density estimation using multiple energy band information.