Deviation Threshold Research Articles

Regional Symmetry of the Pelvis

Understanding the left-right symmetry of the pelvis can assist with the virtual reconstruction of fractured pelvises in the future. Further specifying the different regions can give more information regarding the anatomy of the bone. CT scans of intact pelvises are turned into 3D models using a computer program called Mimics. The femurs and spine are removed manually to isolate the bone. The model can then be analyzed and the regions defined. The pelvis is split into three pieces using the program Geomagic, the iliac, acetabular, and pubic rami regions. The two cuts are made based around the highest and lowest points of the acetabular cup. From here each cut piece can be mirrored and aligned with the opposing side of the pelvis. Colour deviation maps can then be generated to visualize the asymmetry, as well as the RMS value and the percentage of points within a 2 mm deviation threshold. The average RMS values are below 2 mm and the percentage of points within 2 mm is high. My research reflects that the pelvis is symmetrical and may be used to assist in the surgical planning process of pelvic fractures.

Quantification of intramyocardial hemorrhage volume using magnetic resonance imaging with three-dimensional T1-weighted sequence in patients with ischemia-reperfusion injury: a semi-automated image processing technique.

Although intramyocardial hemorrhage (IMH) is a poor prognostic factor caused by ischemia reperfusion injury, little evidence is available regarding the association between IMH volume and biomarkers. In the present study, we measured IMH volume using three-dimensional (3D) T1-weighted magnetic resonance imaging (T1-MRI) and investigated its association with biomarkers. Moreover, the accuracy of semi-automatic measurement of IMH volume was validated. We retrospectively enrolled 33 consecutive patients (mean age 67 ± 11years) who underwent cardiac MRI after reperfusion therapy for acute myocardial infarction. IMH was observed in 4 patients (12.1%). Receiver operating characteristics (ROC) analysis of creatine kinase (CK) and CK-muscle/brain (CK-MB) tests for detecting IMH were performed. IMH volume measured using semi-automatic methods by a 2 standard deviation (SD) threshold was compared to manual measurements using the Spearman's correlation coefficient (ρ) and Bland-Altman analyses. ROC analysis revealed optimal cutoff values of CK: 2460IU/l and CK-MB: 231IU/l (area under the curve: 0.95 and 0.91; sensitivity: 86% and 79%; specificity: 100% for both). IMH volume with the 2SD threshold correlated with that of the manual measurement [5.84g (3.30 to 9.00)g vs. 8.07g (5.37 to 9.33); ρ: 0.85, p < 0.01; bias (limit of agreement): -0.01g (-0.51 to 0.49); intraclass correlation coefficients 0.84 (0.75 to 0.90)]. Our findings could help identify the risk of IMH after reperfusion therapy with biomarkers. 3D T1-MRI can semi-automatically provide accurate IMH volume without being time-consuming.

Sedentary Thresholds for Accelerometry-Based Mean Amplitude Deviation and Electromyography Amplitude in 7-11 Years Old Children.

We investigated the ability of energy expenditure, movement sensing, and muscle activity to discriminate sedentary and non-sedentary activities in children. Thirty-five 7–11-year-old children participated in the study. Simultaneous assessment of oxygen uptake (V̇O2), triaxial accelerometry, and thigh muscle electromyography (EMG) were performed during eight different sedentary and non-sedentary activities including lying down, sitting-, standing-, and walking-related activities, which were performed in a random order. Mean values of V̇O2, accelerometry, and EMG from the concurrent 2 min epochs during each activity were computed. Resting energy expenditure (REE) was measured during 30 min supine rest. Directly measured metabolic equivalent of tasks (METs, V̇O2 in activities/V̇O2 in REE) were calculated for each activity. Mean amplitude deviation (MAD) was computed for accelerometry. EMG was normalized for mean muscle activity during self-paced walking. The classification accuracy of METs, MAD, and EMG to discriminate sedentary activities from physical activities was investigated by receiver operating characteristic curves and optimal cut-offs based on maximal sensitivity and specificity. Mean (SD) REE was 5.0 ± 0.8 ml/kg/min. MET, MAD, and EMG values ranged from 1.0 to 4.9, 0.0020 to 0.4146 g, and 4.3 to 133.9% during lying down and walking at 6 km/h, respectively. Optimal cut-offs to discriminate sedentary activities from non-sedentary activities were 1.3 for METs (sensitivity = 82%, specificity = 88%), 0.0033 g for MAD (sensitivity = 80%, specificity = 91%), and 11.9% for EMG (sensitivity = 79%, specificity = 92%). In conclusion, this study provides applicable thresholds to differentiate sitting and standing and sedentary and non-sedentary activities based on METs, MAD, and EMG in young children.

Passive mixing rate of trapped squeezed nanodroplets—A time scale analysis

The elegance of digital microfluidics is in incorporating high quantities of manipulated micro and nanodroplets on-chip, each of which can be considered a small-volume carrier of various chemical and biological reagents. Therefore, the analysis of on-demand manipulation of these micro and nanocarriers is extremely important in developing an optimized lab on a chip. In this work, passive coalescence and mixing between two trapped, squeezed nanodroplets inside a closed microfluidic device was investigated. The droplets are composed of glycerol dyed in blue and red, dispersed inside oleic acid as the carrier oil. A microwell with a circular cross section was fabricated on the top wall of the microchannels to trap the first droplet for increasing the mixing precision and minimizing the viscous shear stress imposed on the droplets from the channel walls. The energy minimization theory was used to develop a parametric study for this trapping technique and to choose the optimum design parameters for droplet trapping in terms of efficiency. Image processing was performed on the snapshots of the trapped glycerol nanodroplets during mixing. Growth in passive mixing percentage was demonstrated to be asymptotical and was formulated with an empirical equation of exponential form as a function of the passive mixing relaxation time. The required time for the passive mixing of a glycerol droplet pair was measured considering various thresholds for the final standard deviation of the gray intensity indices. This finding was of the order of magnitude of the diffusive mixing time scale and physically consistent with the Stokes flow regime.

&lt;p&gt;Quantifying choriocapillaris hypoperfusion in patients with choroidal neovascularization using swept-source OCT angiography&lt;/p&gt;

PurposeTo compare choriocapillaris flow voids (FV) in patients with neovascular age-related macular degeneration (AMD) with age-matched normal controls using swept-source optical coherence tomography angiography (SS-OCTA).Patients and methodsEleven eyes of 11 subjects with neovascular AMD and 11 eyes of 11 age-similar normal subjects were imaged using SS-OCTA with a 6x6mm scanning pattern. Choriocapillaris FV, defined as a percentage of regions determined to have flow deficits divided by the total scanned region, was measured using a one standard deviation thresholding algorithm developed from a database of age-similar normal subjects.ResultsChoriocapillaris FV was more extensive in patients with choroidal neovascularization (CNV) compared to age-similar normal subjects (FV: 20.56±4.95, 95% CI: 17.64–23.49 vs FV: 10.95±2.08, 95% CI: 9.73–12.18, respectively; P=0.0001). FV within a two-degree margin surrounding CNV in wet AMD subjects (FV: 35.04±9.34; 95% CI: 29.52–40.56) was increased compared to normal subjects (P<0.001). FV of the region outside the two-degree margin surrounding CNV (FV: 19.61±6.08, 95% CI: 16.02–23.20) was increased compared to age-similar controls (P=0.0002). In neovascular AMD eyes, FV was greater within two degrees of the margin of CNV than in the remainder of the macula (margin: 35.04±9.34; outside: 19.61±6.08; P=0.002), and CNV lesion area was positively correlated with FV (correlation coefficient =0.84; 95% CI: 0.49–0.96; P=0.001).ConclusionChoriocapillaris flow deficits were significantly greater in wet AMD subjects than age-similar normal subjects, suggesting that choroidal hypo-perfusion likely plays a role in the pathogenesis of neovascular AMD. Recognition of choriocapillaris flow deficits in patients with AMD may facilitate earlier diagnosis and identify alternative therapeutic targets for this multifactorial disease.

A Hybrid Model Based on Principal Component Analysis, Wavelet Transform, and Extreme Learning Machine Optimized by Bat Algorithm for Daily Solar Radiation Forecasting

Precise solar radiation forecasting is of great importance for solar energy utilization and its integration into the grid, but because of the daily solar radiation’s intrinsic non-stationary and nonlinearity, which is influenced by a lot of elements, single predicting models may have difficulty obtaining results with high accuracy. Therefore, this paper innovatively puts forward an original hybrid model that predicts solar radiation through extreme learning machine (ELM) optimized by the bat algorithm (BA) based on wavelet transform (WT) and principal component analysis (PCA). First, choose the meteorological variables on the basis of Pearson coefficient test, and WT will decompose historical solar radiation into two time series, which are de-noised signal and noise signal. In the approximate series, the lag phase of historical radiation is obtained by partial autocorrelation function (PACF). After that, use PCA to reduce the dimensions of the influencing factors, including meteorological variables and historical radiation. Finally, ELM is established to predict daily solar radiation, whose input weight and deviation thresholds gained optimization by BA, thus it is called BA-ELM henceforth. In view of the four distinct solar radiation series obtained by NASA, the empirical simulation explained the hybrid model’s validity and effectiveness compared to other primary methods.

Along‐Slope Variability of Cross‐Slope Eddy Transport in East Antarctica

AbstractCircumpolar Deep Water (CDW) transport across the Antarctic continental slope regulates the delivery of heat to the shelf and its availability to melt floating ice shelves. The cross‐slope density field, calculated from profiles collected by conductivity‐temperature‐depth‐tagged marine mammals on the East Antarctic slope (0–160°E, above 1,000‐ to 3,000‐m isobaths), indicates eddy‐driven overturning: onshore transport of CDW and offshore transport of shallower Antarctic Surface Water. Enhanced eddy activity, determined by a spice standard deviation threshold in the CDW layer, is present over about a third of the East Antarctic slope analyzed. Significantly stronger CDW transport in regions of elevated spice variability produces subsurface temperature anomalies of 0.2–0.25 °C relative to the East Antarctic average. Estimating eddy diffusivity from the hydrography yields about 0.8 m2/s of warm CDW transport to the shelf break in high‐variability regions. Variability of eddy‐induced CDW transport influences the reservoir of heat available for transport across the shelf break.

Realtime Indoor Workout Analysis Using Machine Learning & Computer Vision.

These days there are thousands of workout videos available on the internet. Samsung Health2 [1] provides a dedicated section called programs containing short workout videos for various exercises. The goal is to assist people perform these workouts independently on their own. A common observation is that even people who visit gym regularly find it difficult to perform all steps (body pose alignments) in a workout accurately. Continuously doing an exercise incorrectly may eventually cause severe long term injuries. To help solve this problem and provide assistance in form of a visual feedback while performing a workout, we propose a system to analyze user's body posture during a workout and compare it to a professional's reference workout. We represent human body as a collection of limbs and analyze angle between limb pairs to detect errors and provide corrective action to the user. Our system builds on the latest advancements using deep learning for human body pose estimation. We use techniques for time series data alignment like DTW [2] (Dynamic Time Warping) along with optical flow tracking to synchronize user/reference videos. We are able to detect and locate errors in user's activity (pose) very effectively based on some threshold deviation between the limb angles. The system in future can be extended to be used by physicians to monitor patient's recovery following an injury.

Cascade interpolation learning with double subspaces and confidence disturbance for imbalanced problems

In this paper, a new ensemble framework named Cascade Interpolation Learning with Double subspaces and Confidence disturbance (CILDC) is designed for the imbalanced classification problems. Developed from the Cascade Forest of the Deep Forest which is the stacking based tree ensembles for big data issues with less hyper-parameters, CILDC aims to generalize the cascade model for more base classifiers. Specifically, CILDC integrates base classifiers through the double subspaces strategy and the random under-sampling preprocessing. Further, one simple but effective confidence disturbance technique is introduced to CILDC to tune the threshold deviation for imbalanced samples. In detail, the disturbance coefficients are multiplied to various confidence vectors before interpolating in each level of CILDC, and the ideal threshold can be adaptively learned through the cascade structure. Furthermore, both the Random Forest and the Naive Bayes are suitable to be the base classifier for CILDC. Subsequently, comprehensive comparison experiments on typical imbalanced datasets demonstrate both the effectiveness and generalization of CILDC.

Long-term robust identification potential of a wavelet packet decomposition based recursive drift correction of E-nose data for Chinese spirits

The drift of electronic nose (E-nose) is always yielded, and makes it does not possess long-term robust detection ability, so that the detection accuracy of the same samples tested in the subsequent period will be reduced. In order to enhance the long-term identification robustness of six kinds of Chinese spirits, a recursive identification model was established. Firstly, E-nose data were decomposed by a wavelet packet and generated decomposition coefficients. Then a relative deviation threshold function was constructed to handle these coefficients. And then, the E-nose data with little drift or not were obtained by reconstructing the corrected coefficients. Finally, a concept of “sample test time window” (SMTW) was introduced for building the recursive identification model. The six kinds of spirit samples were discontinuously tested for 16 months, and the SMTW was determined as 6 months. As SMTW moves forward for 2 months every time, the recursive identification model based on Fisher discriminant analysis (FDA) was also built and the correct identification rate was 96.5%, namely the tested samples in 2 months of following SMTW could be accurately identified. This illustrates that the proposed methods are very effective.

Predictive Models to Determine Clinically Relevant Deviations in Delivered Dose for Head and Neck Cancer

PurposeThis study aimed to improve the understanding of deviations between planned and accumulated doses and to establish metrics to predict clinically significant dosimetric deviations midway through treatment to evaluate the potential need to re-plan during fractionated radiation therapy (RT). Methods and materialsA total of 100 patients with head and neck cancer were retrospectively evaluated. Contours were mapped from the planning computed tomography (CT) scan to each fraction cone beam CT via deformable image registration. The dose was calculated on each cone beam CT and evaluated based on the mapped contours. The mean dose at each fraction was averaged to approximate the accumulated dose for structures with mean dose constraints, and the daily maximum dose was summed to approximate the accumulated dose for structures with maximum dose constraints. A threshold deviation value was calculated to predict for patients needing midtreatment re-planning. This predictive model was applied to 52 patients treated at a separate institution. ResultsDose was accumulated on 10 organs over 100 patients. To generate a threshold deviation that predicted the need to re-plan with 100% sensitivity, the submandibular glands required re-planning if the delivered dose was at least 3.5 Gy higher than planned by fraction 15. This model predicts the need to re-plan the submandibular glands with 98.7% specificity. In the independent evaluation cohort, this model predicts the need to re-plan the submandibular glands with 100% sensitivity and 98.0% specificity. The oral cavity, intermediate clinical target volume, left parotid, and inferior constrictor patient groups each had 1 patient who exceeded the threshold deviation by the end of RT. By fraction 15 of 30 to 35 total fractions, the left parotid gland, inferior constrictor, and intermediate clinical target volume had a dose deviation of 3.1 Gy, 5.9 Gy, and 4.8 Gy, respectively. When a deformable image registration failure was observed, the dose deviation exceeded the threshold for at least 1 organ, demonstrating that an automated deformable image registration–based dose assessment process could be developed with user evaluation for cases that result in dose deviations. ConclusionsA midtreatment threshold deviation was determined to predict the need to replan for the submandibular glands by fraction 15 of 30 to 35 total fractions of RT.

Performance comparison of automatic peak detection for signal analyser

The aim of this paper is to propose a new peak detection method for a portable device, which know as modified automatic threshold peak detection (M-ATPD). M-ATPD evolves out of ATPD with a focus on reducing computational time. The proposed method replaces the clustering threshold calculation in ATPD with a standard deviation threshold calculation. M-ATPD reduces computational time by 2 times faster compared to ATPD for control signal and 8.65 times faster compared to ATPD for raw biosignals. Modified ATPD also shows a slight improvement in terms of detection error, with a decrease of about 6.66% to 13.33% in peak detection of noise signals. Modified ATPD successfully fixes the error of peak detection on pulse control signals associated with ATPD. For raw biosignals, in total M-ATPD achieved 19.41% lower detection error compare to ATPD.

Research on an Auto-Optimized Capacitor Voltage Balancing Control Strategy of MMC SM for Renewable Energy HVDC Transmission System

The Modular Multilevel Converter (MMC) is one of the most attractive converter topologies in the High Voltage Direct Current (HVDC) transmission field. The latest widely used sorting method has a low algorithm complexity. It can effectively balance the sub-module (SM) capacitor voltages, but it would cause relatively high switching frequency and power loss. Aiming at the problem that the sub-module (SM) capacitor voltage sorting algorithm has a large switching loss due to the high switching frequency of the device, this paper proposes an auto-optimized capacitor voltage balancing control strategy. Firstly, the topology and operation principle of MMC are analyzed. Secondly, a SM capacitor voltage control method based on the dynamic deviation threshold is proposed. Considering the switch switching state of the SM and the difference between the voltages of each SM, the algorithm can obtain the dynamic deviation valve using the closed-loop control. The method can avoid the unnecessary repeated switching of the Insulated Gate Bipolar Transistor (IGBT) under the premise of ensuring that the capacitance voltages of the SMs are basically the same, which effectively result in reducing the switching frequency of the MMC SM and reducing the switching loss, thereby improving the operating efficiency of the system. Finally aiming at the proposed control strategy, the simulation and experimental verification are carried out which shows that the proposed algorithm can better control the system voltage deviation, reduce the switching loss of the system and improve the stability of the system.

A novel parameter derived from photoplethysmographic pulse wave to distinguish preeclampsia from non-preeclampsia.

To describe the comparative hierarchical area ratio (CHAR), a novel parameter derived from the photoplethysmographic (PPG) pulse wave and differences in CHAR values in parturients with and without preeclampsia (PE). A total of 59 parturients (37 without and 22 with PE) was conducted at the Women's Hospital, Zhejiang University, School of Medicine in Hangzhou, China. We calculated the CHAR values derived from the PPG pulse wave and compared them in parturients with and without PE. The values of CHAR derived from the parturients with PE were lower compared to those without PE (p < 0.01). The ROC analysis indicated that the best threshold for the mean value of CHAR was 7.92 to predict PE with a sensitivity of 86.4% and a specificity of 87.1%, while the threshold for the standard deviation of CHAR was 0.76 with a sensitivity of 77.3% and a specificity of 77.4%. The area under the curve (AUC) was 0.91 for mean value of CHAR while 0.78 for standard deviation of CHAR. Meanwhile, a contrast of AUC between CHAR and the former parameter we proposed showed CHAR had better performance in distinguishing PE (0.908 over 0.615, p < 0.01). The novel parameter, CHAR, derived from PPG pulse wave differs in parturients with and without PE with high sensitivity and specificity, suggesting that the CHAR might be an effective tool in differentiating the presence of PE.

Model Based Diagnosis of Oxygen Sensors

Abstract Automotive industry targets such as complying with emission legislations and increasing fuel economy, require the improvement of air-fuel ratio control systems. Oxygen sensors are a crucial part of these control systems and regulations oblige monitoring of their performance and detecting sensor-related faults. The primary purpose of this paper is to develop a methodology for precise and accurate monitoring and diagnosis of oxygen sensors to meet legislations and performance targets while the required calibration effort is reduced. Input parameters with the highest correlation factors were selected to be utilized in different system identification methodologies to statistically determine the most fitting model. In the end, a NARX model with two hidden layers and eight neurons in each hidden layer with standard deviation and mean threshold values was determined to be the optimum design to detect if the oxygen sensor was functioning or faulty.

Macular evaluation of the retinal and choroidal vasculature changes in anterior ischemic optic neuropathy-a case control study

BackgroundThis study aimes to characterize the fundus structural changes in patients with nonarteritic anterior ischemic optic neuropathy (NAION) and the correlation between macular vessel density, retinal nerve fibre layer (RNFL) parameters and visual field sensitivity (VFS) in NAION patients.MethodsA retrospective case control study was performed using 37 eyes with NAION, 30 uninvolved contralateral eyes, and 27 eyes of healthy age-matched subjects. Data on the retinas and choroidal vessel densities and VFS were compared among the three groups.ResultsThe NAION group exhibited significantly lower RNFL thicknesses, lower ganglion cell complexes (GCC), larger global loss volume (GLV) values and focal loss volume (FLV) values when compared with both uninvolved eyes and healthy eyes (p < 0.01 for all comparisons). The superficial vessel density (SVD) valus (whole, parafovea, superior-hemi and inferior-hemi) were significantly lower in NAION eyes, followed by uninvolved eyes and healthy eyes (p < 0.01; LSD, p < 0.05 for all comparisons). The deep vessel density (DVD) values (parafovea, superior-hemi and inferior-hemi) were the lowest by a significant value in NAION eyes, followed by uninvolved eyes and healthy eyes (p < 0.01; LSD, p < 0.05 for all comparisons). However, DVD value measurements (whole and fovea) of healthy and uninvolved eyes were not significantly different. The average threshold deviation (TD) was − 11.02 ± 3.75 dB for the overall field region, − 6.01 ± 2.21 dB for the affected superior field region and − 9.98 ± 3.34 dB for the affected inferior field region in NAION eyes. A statistically significant correlation was found between the RNFL thickness and visual field(VF) loss (r = − 0.788, p < 0.001).ConclusionIn addition to peripapillary vascular changes occurring in NAION eyes, macular vessel density is also involved. Furthermore, NAION-uninvolved eyes exhibited abnormalities compared with healthy eyes. This indicates that vascular changes may occur before changes in retinal thickness at the early stages of NAION.

A 5 g Inertial Micro-Switch with Enhanced Threshold Accuracy Using Squeeze-Film Damping.

Our previous report based on a 10 g (gravity) silicon-based inertial micro-switch showed that the contact effect between the two electrodes can be improved by squeeze-film damping. As an extended study toward its potential applications, the switch with a large proof mass suspended by four flexible serpentine springs was redesigned to achieve 5 g threshold value and enhanced threshold accuracy. The impact of the squeeze-film damping on the threshold value was theoretically studied. The theoretical results show that the threshold variation from the designed value due to fabrication errors can be reduced by optimizing the device thickness (the thickness of the proof mass and springs) and then establishing a tradeoff between the damping and elastic forces, thus improving the threshold accuracy. The design strategy was verified by FEM (finite-element-method) simulation and an experimental test. The simulation results show that the maximum threshold deviation was only 0.15 g, when the device thickness variation range was 16–24 μm, which is an adequately wide latitude for the current bulk silicon micromachining technology. The measured threshold values were 4.9–5.8 g and the device thicknesses were 18.2–22.5 μm, agreeing well with the simulation results. The measured contact time was 50 μs which is also in good agreement with our previous work.

Using 1 HN amide temperature coefficients to define intrinsically disordered regions: An alternative NMR method.

This report describes a cost-effective experimental method for determining an intrinsically disordered protein (IDP) region in a given protein sample. In this area, the most popular (and conventional) means is using the amide (1 HN ) NMR signal chemical shift distributed in the range of 7.5-8.5 ppm. For this study, we applied an additional step: analysis of 1 HN chemical shift temperature coefficients (1 HN -CSTCs) of the signals. We measured 1 H-15 N two-dimensional NMR spectra of model IDP samples and ordered samples at four temperatures (288, 293, 298, and 303 K). We derived the 1 HN -CSTC threshold deviation, which gives the best correlation of ordered and disordered regions among the proteins examined (below -3.6 ppb/K). By combining these criteria with the newly optimized chemical shift range (7.8-8.5 ppm), the ratios of both true positive and true negative were improved by approximately 19% (62-81%) compared with the conventional "chemical shift-only" method.

Feature based transition region extraction for image segmentation: Application to worm separation from leaves

Transition region based approaches are recent hybrid segmentation techniques well known for its simplicity and effectiveness. Here, the segmentation effectiveness depends on robust extraction of transition regions. So, we have proposed transition region extraction method for image segmentation. The proposed method initially decomposes the gray image in wavelet domain. Local standard deviation filtering and thresholding operation is used to extract transition region feature matrix. Using this feature matrix, the corresponding prominent wavelet coefficients of different bands are found. The inverse wavelet transform is then applied to the modified coefficients to get edge image with more than one-pixel width. Global thresholding is applied to get transition regions. Further, it undergoes morphological thinning and region filling operation to extract the object regions. Finally, the objects are extracted using the object regions. The proposed method is compared with different image segmentation methods. An experimental result reveals that the proposed method outperforms other methods for segmentation of images containing single and multiple objects. The proposed method can also be applied for worm separation from leaves.

An improved method to render the sound of VR scene

Based on the virtual scene containing hundreds of movable sound sources, due to the high computational cost of clustering stage, the traditional spatial sound rendering schemes often take up too much computing resources, which have become a bottleneck in the development of VR audio rendering technology. In this paper, we use fractional Fourier transform (FRFT) as a tool in sound sampling to reduce the quantization noise during the ADC conversion stage. Moreover, we improve the processing speed of sound rendering and the operation efficiency of the entire system by adding the average angle deviation threshold in the clustering step. In addition, we design and implement a perceptual user experiment, and validates the notion that people are more susceptible to spatial errors in different types of sound sources, especially if it is visible. Based on this conclusion, this paper proposes an improved method of sound clustering, which reduces the possibility of clustering different types of sound sources.