Pub Date : 2019-07-01DOI: 10.1109/SSPD.2019.8751653
Leon Kocjančič, A. Balleri, T. Merlet
This paper presents an analysis of the Doppler tolerance and isolation properties of five different sets of piecewise linear frequency modulated (PLFM) waveform triplets consisting of a combination of LFM subchirps. Different combinations of PLFM signals are used to produce waveforms with the same time-bandwidth product and optimise them with respect to isolation. The performance of the proposed waveforms are numerically investigated and a comparison between sets is presented. Results confirm that the waveforms have quasi-orthogonal properties and exhibit a degree of Doppler tolerance.
{"title":"Numerical Characterisation of Quasi-Orthogonal Piecewise Linear Frequency Modulated Waveforms","authors":"Leon Kocjančič, A. Balleri, T. Merlet","doi":"10.1109/SSPD.2019.8751653","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751653","url":null,"abstract":"This paper presents an analysis of the Doppler tolerance and isolation properties of five different sets of piecewise linear frequency modulated (PLFM) waveform triplets consisting of a combination of LFM subchirps. Different combinations of PLFM signals are used to produce waveforms with the same time-bandwidth product and optimise them with respect to isolation. The performance of the proposed waveforms are numerically investigated and a comparison between sets is presented. Results confirm that the waveforms have quasi-orthogonal properties and exhibit a degree of Doppler tolerance.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126918669","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/SSPD.2019.8751664
Abderrahim Halimi, Rachael Tobin, A. Mccarthy, J. Bioucas-Dias, S. Mclaughlin, G. Buller
The aim of this paper is to propose a specialized algorithm to process Multitemporal or Multispectral 3D single-photon Lidar images. Of particular interest are challenging scenarios often encountered in real world, i.e., imaging through obscurants such as water, fog or imaging multilayered targets such as target behind camouflage. To restore the data, the algorithm accounts for data Poisson statistics and available prior knowledge regarding target depth and reflectivity estimates. More precisely, it accounts for (a) the non-local spatial correlations between pixels, (b) the spatial clustering of target returned photons and (c) spectral and temporal correlations between frames. An alternating direction method of multipliers (ADMM) algorithm is used to minimize the resulting cost function since it offers good convergence properties. The algorithm is validated on real data which show the benefit of the proposed strategy especially when dealing with multi-dimensional 3D data.
{"title":"Joint Reconstruction of Multitemporal or Multispectral Single-Photon 3D LiDAR Images","authors":"Abderrahim Halimi, Rachael Tobin, A. Mccarthy, J. Bioucas-Dias, S. Mclaughlin, G. Buller","doi":"10.1109/SSPD.2019.8751664","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751664","url":null,"abstract":"The aim of this paper is to propose a specialized algorithm to process Multitemporal or Multispectral 3D single-photon Lidar images. Of particular interest are challenging scenarios often encountered in real world, i.e., imaging through obscurants such as water, fog or imaging multilayered targets such as target behind camouflage. To restore the data, the algorithm accounts for data Poisson statistics and available prior knowledge regarding target depth and reflectivity estimates. More precisely, it accounts for (a) the non-local spatial correlations between pixels, (b) the spatial clustering of target returned photons and (c) spectral and temporal correlations between frames. An alternating direction method of multipliers (ADMM) algorithm is used to minimize the resulting cost function since it offers good convergence properties. The algorithm is validated on real data which show the benefit of the proposed strategy especially when dealing with multi-dimensional 3D data.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133898208","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-07-01DOI: 10.1109/SSPD.2019.8751640
L. Spyrou, P. Chambers, M. Sellathurai, J. Thompson
Wireless sensor networks enjoy many advantages over wired networks due to their ability to be deployed easily and flexibly in many scenarios. However, they suffer from the drawback that the environment may be unknown and hence the sensor network cannot easily be optimised for it. Furthermore, state-of-the-art studies only consider the detection and localisation performance separately. The main novelty of this work is that we compare the theoretical properties of three scanning strategies both in terms of their detection and localisation performance. We consider: a) sequential scanning, where all sensors scan the channels in sequence, b) groupwise scanning, where the sensors are split into groups with each one performing a sequential scan, and c) random scanning, where each sensor is assigned a channel at random. We demonstrate the theoretical properties of the strategies and perform a numerical evaluation for a typical radio surveillance scenario. The tradeoffs of the methods between detection and localisation performance are demonstrated to be dependent on the detection and localisation accuracy with respect to the number of sensors. Approximate knowledge of those curves can aid in the design of an optimal sensor scanning strategy.
{"title":"Tradeoffs in Detection and Localisation Performance for Mobile Sensor Scanning Strategies","authors":"L. Spyrou, P. Chambers, M. Sellathurai, J. Thompson","doi":"10.1109/SSPD.2019.8751640","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751640","url":null,"abstract":"Wireless sensor networks enjoy many advantages over wired networks due to their ability to be deployed easily and flexibly in many scenarios. However, they suffer from the drawback that the environment may be unknown and hence the sensor network cannot easily be optimised for it. Furthermore, state-of-the-art studies only consider the detection and localisation performance separately. The main novelty of this work is that we compare the theoretical properties of three scanning strategies both in terms of their detection and localisation performance. We consider: a) sequential scanning, where all sensors scan the channels in sequence, b) groupwise scanning, where the sensors are split into groups with each one performing a sequential scan, and c) random scanning, where each sensor is assigned a channel at random. We demonstrate the theoretical properties of the strategies and perform a numerical evaluation for a typical radio surveillance scenario. The tradeoffs of the methods between detection and localisation performance are demonstrated to be dependent on the detection and localisation accuracy with respect to the number of sensors. Approximate knowledge of those curves can aid in the design of an optimal sensor scanning strategy.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"74 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121899547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-05-17DOI: 10.1109/SSPD.2019.8751641
Raied Caromi, M. Souryal
In the 3.5 GHz Citizens Broadband Radio Service (CBRS), 100 MHz of spectrum will be dynamically shared between commercial users and federal incumbents. Dynamic use of the band relies on a network of sensors dedicated to detecting the presence of federal incumbent signals and triggering protection mechanisms when necessary. This paper uses field-measured waveforms of incumbent signals in and adjacent to the band to evaluate the performance of support vector machine (SVM) classifiers for these sensors. We find that a peak analysis classifier and a higher-order statistics classifier perform comparably when the signal is in white Gaussian noise or commercial long term evolution (LTE) emissions, but with out-of-band emissions of adjacent-band systems the peak analysis classifier is far superior. This result also highlights the importance of including adjacent-band emissions in any performance evaluation of 3.5 GHz sensors.
{"title":"Detection of Incumbent Radar in the 3.5 GHz CBRS Band using Support Vector Machines","authors":"Raied Caromi, M. Souryal","doi":"10.1109/SSPD.2019.8751641","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751641","url":null,"abstract":"In the 3.5 GHz Citizens Broadband Radio Service (CBRS), 100 MHz of spectrum will be dynamically shared between commercial users and federal incumbents. Dynamic use of the band relies on a network of sensors dedicated to detecting the presence of federal incumbent signals and triggering protection mechanisms when necessary. This paper uses field-measured waveforms of incumbent signals in and adjacent to the band to evaluate the performance of support vector machine (SVM) classifiers for these sensors. We find that a peak analysis classifier and a higher-order statistics classifier perform comparably when the signal is in white Gaussian noise or commercial long term evolution (LTE) emissions, but with out-of-band emissions of adjacent-band systems the peak analysis classifier is far superior. This result also highlights the importance of including adjacent-band emissions in any performance evaluation of 3.5 GHz sensors.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129099570","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-05-09DOI: 10.1109/SSPD.2019.8751643
D. J. Sadler
Adcock Watson-Watt (AWW) methods for radio direction finding (DF) have a long history, but are often still used in modern, wideband DF systems. A number of sources of error can reduce the accuracy of the DF estimates produced. One such source of error is due to unaccounted for amplitude and phase errors in the three receiver channels; perhaps due to unbalance in the analogue circuitry, or component tolerance in the receive filters. This paper provides a theoretical analysis of the impact of complex gain errors on the expected DF estimates. When combined with the effects of receiver noise, analytical performance curves for the AWW system can be produced. For small array apertures, it is shown that AWW can actually outperform more sophisticated N-channel DF such as correlative interferometer, maximum likelihood and subspace techniques.
{"title":"Accuracy of Adcock Watson-Watt DF in the Presence of Channel Errors","authors":"D. J. Sadler","doi":"10.1109/SSPD.2019.8751643","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751643","url":null,"abstract":"Adcock Watson-Watt (AWW) methods for radio direction finding (DF) have a long history, but are often still used in modern, wideband DF systems. A number of sources of error can reduce the accuracy of the DF estimates produced. One such source of error is due to unaccounted for amplitude and phase errors in the three receiver channels; perhaps due to unbalance in the analogue circuitry, or component tolerance in the receive filters. This paper provides a theoretical analysis of the impact of complex gain errors on the expected DF estimates. When combined with the effects of receiver noise, analytical performance curves for the AWW system can be produced. For small array apertures, it is shown that AWW can actually outperform more sophisticated N-channel DF such as correlative interferometer, maximum likelihood and subspace techniques.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125398140","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-05-09DOI: 10.1109/SSPD.2019.8751646
D. Parker, Henry White, J. Oakley, G. Bishop
For new military aircraft a specification of sensor characteristics and performance is required at an early stage in the design cycle, well before testing of a prototype. In the early days of military aviation the Johnson criteria [1] were used to determine the sensor resolution required for target recognition by a human. In the present day sensor data are processed by computer using various Automatic Target Recognition (ATR) algorithms. However there is no accepted method for predicting the sensor resolution and SNR required for reliable ATR and hence there is risk that any chosen sensor may fail to support the required ATR performance. This paper reports a study into the use of publicly-available CAD models for aircraft to address this requirement. The study considers the worst-case confusion between two views of 15 different aircraft types. For simplicity only rotations by an angle θ about the Z (vertical) axis are considered. Firstly the sensor resolution and noise level is fixed. Then for each aircraft type and view angle an ensemble of synthetic silhouettes are generated. Using these ensembles, a-posteriori distributions of 5 standard scale-invariant shape features (eccentricity, orientation, solidity, circularity and bounding box aspect ratio) are estimated for each view angle θ. The performance of ATR at the given resolution and noise level is predicted by estimating the Bayes Error Rate [2] when deciding between each aircraft type and the 14 non-matching types using these features. The worst-case confusion in terms of erroneous aircraft type and view angle is identified. The sensor resolution is then changed and the above process repeated to investigate the effect of varying sensor resolution on performance. As expected, high sensor resolution leads to low probability of misclassification, even in the worst-case. Reduction in resolution and increasing noise level causes the Bayes Error Rate to rise quickly. The Bayes Error Rate gives a fundamental limit to the reliability of classification, irrespective of the actual type of classification algorithm used. The predictions from the model are confirmed by testing against a standard classifier for specific discrimination examples. Further development of the approach presented is expected to yield a method for specifying sensor resolution requirements for specific ATR problems.
{"title":"Prediction of Sensor Performance Required for Reliable Aircraft Target Discrimination","authors":"D. Parker, Henry White, J. Oakley, G. Bishop","doi":"10.1109/SSPD.2019.8751646","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751646","url":null,"abstract":"For new military aircraft a specification of sensor characteristics and performance is required at an early stage in the design cycle, well before testing of a prototype. In the early days of military aviation the Johnson criteria [1] were used to determine the sensor resolution required for target recognition by a human. In the present day sensor data are processed by computer using various Automatic Target Recognition (ATR) algorithms. However there is no accepted method for predicting the sensor resolution and SNR required for reliable ATR and hence there is risk that any chosen sensor may fail to support the required ATR performance. This paper reports a study into the use of publicly-available CAD models for aircraft to address this requirement. The study considers the worst-case confusion between two views of 15 different aircraft types. For simplicity only rotations by an angle θ about the Z (vertical) axis are considered. Firstly the sensor resolution and noise level is fixed. Then for each aircraft type and view angle an ensemble of synthetic silhouettes are generated. Using these ensembles, a-posteriori distributions of 5 standard scale-invariant shape features (eccentricity, orientation, solidity, circularity and bounding box aspect ratio) are estimated for each view angle θ. The performance of ATR at the given resolution and noise level is predicted by estimating the Bayes Error Rate [2] when deciding between each aircraft type and the 14 non-matching types using these features. The worst-case confusion in terms of erroneous aircraft type and view angle is identified. The sensor resolution is then changed and the above process repeated to investigate the effect of varying sensor resolution on performance. As expected, high sensor resolution leads to low probability of misclassification, even in the worst-case. Reduction in resolution and increasing noise level causes the Bayes Error Rate to rise quickly. The Bayes Error Rate gives a fundamental limit to the reliability of classification, irrespective of the actual type of classification algorithm used. The predictions from the model are confirmed by testing against a standard classifier for specific discrimination examples. Further development of the approach presented is expected to yield a method for specifying sensor resolution requirements for specific ATR problems.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121042332","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-05-09DOI: 10.1109/SSPD.2019.8751651
D. Vint, G. Di Caterina, J. Soraghan, R. Lamb, D. Humphreys
This paper presents an evaluation of the suitability of the Very Deep Super Resolution (VDSR) architecture, to increase the spatial resolution of lower quality images. For this aim, two sets of tests are performed. The former being on real life images to determine the networks ability to improve low resolution images. The second test is performed on images of a resolution chart, and therefore synthetic. This is to analyse the frequency response of the network. For each test, three metrics are used to assess image quality. These are the Peak Signal to Noise Ratio (PSNR), Structural Similarity Index (SSIM) and Modulation Transfer Function (MTF). Experimental results show that the VDSR network is able to increase the quality of the images within the first test in all three metrics, therefore showing that the network is suitable for super resolution. The second test provides more information on the limitations of the network when given a high contrast image, and the resulting ringing effects it can create. Therefore results in PSNR/SSIM values are not improved over the low resolution images, however they have a higher MTF curve as well as more visually sharp images.
{"title":"Evaluation of Performance of VDSR Super Resolution on Real and Synthetic Images","authors":"D. Vint, G. Di Caterina, J. Soraghan, R. Lamb, D. Humphreys","doi":"10.1109/SSPD.2019.8751651","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751651","url":null,"abstract":"This paper presents an evaluation of the suitability of the Very Deep Super Resolution (VDSR) architecture, to increase the spatial resolution of lower quality images. For this aim, two sets of tests are performed. The former being on real life images to determine the networks ability to improve low resolution images. The second test is performed on images of a resolution chart, and therefore synthetic. This is to analyse the frequency response of the network. For each test, three metrics are used to assess image quality. These are the Peak Signal to Noise Ratio (PSNR), Structural Similarity Index (SSIM) and Modulation Transfer Function (MTF). Experimental results show that the VDSR network is able to increase the quality of the images within the first test in all three metrics, therefore showing that the network is suitable for super resolution. The second test provides more information on the limitations of the network when given a high contrast image, and the resulting ringing effects it can create. Therefore results in PSNR/SSIM values are not improved over the low resolution images, however they have a higher MTF curve as well as more visually sharp images.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126438859","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-05-09DOI: 10.1109/SSPD.2019.8751645
D. Cormack, J. Hopgood
Sensor registration is fundamental in sensor fusion. Inaccuracies in sensor location and rotation can manifest themselves into the measurements used in Multiple Target Tracking (MTT), and dramatically degrade its performance. These registration parameters are often estimated separately to any multitarget estimation, which could lead to increased computational expense, and also to systematic errors. Recent works have shown that MTT algorithms derived from Belief Propagation (BP) are computationally efficient and highly scalable for large tracking scenarios. This work presents a hierarchical Bayesian model inspired by single-cluster methods from the Random Finite Set (RFS) literature, that allow for the registration parameters to be estimated jointly with the multiple target tracking. Simulations are carried out on a multistatic radar network containing two radars with a relative range and azimuth bias between them. Results are presented for a particle-BP MTT algorithm, and it's performance is compared to that of a Sequential Monte Carlo (SMC)-Probability Hypothesis Density (PHD) filter. The results show that the BP algorithm outperforms the PHD implementation in terms of accuracy by around 10%.
传感器配准是传感器融合的基础。在多目标跟踪(MTT)中，传感器定位和旋转的不准确性会在测量中表现出来，并显著降低其性能。这些配准参数通常与任何多目标估计分开估计，这可能导致计算费用增加，并且还会导致系统误差。最近的研究表明，基于信念传播(BP)的MTT算法对于大型跟踪场景具有计算效率和高度可扩展性。本文提出了一种受随机有限集(RFS)文献中单聚类方法启发的分层贝叶斯模型，该模型允许在多目标跟踪的同时估计配准参数。在具有相对距离和方位偏差的两台雷达的多基地雷达网络上进行了仿真。给出了粒子- bp MTT算法的结果，并将其性能与序列蒙特卡罗(SMC)-概率假设密度(PHD)滤波进行了比较。结果表明，BP算法的准确率比PHD算法提高了10%左右。
{"title":"Message Passing for Joint Registration and Tracking in Multistatic Radar","authors":"D. Cormack, J. Hopgood","doi":"10.1109/SSPD.2019.8751645","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751645","url":null,"abstract":"Sensor registration is fundamental in sensor fusion. Inaccuracies in sensor location and rotation can manifest themselves into the measurements used in Multiple Target Tracking (MTT), and dramatically degrade its performance. These registration parameters are often estimated separately to any multitarget estimation, which could lead to increased computational expense, and also to systematic errors. Recent works have shown that MTT algorithms derived from Belief Propagation (BP) are computationally efficient and highly scalable for large tracking scenarios. This work presents a hierarchical Bayesian model inspired by single-cluster methods from the Random Finite Set (RFS) literature, that allow for the registration parameters to be estimated jointly with the multiple target tracking. Simulations are carried out on a multistatic radar network containing two radars with a relative range and azimuth bias between them. Results are presented for a particle-BP MTT algorithm, and it's performance is compared to that of a Sequential Monte Carlo (SMC)-Probability Hypothesis Density (PHD) filter. The results show that the BP algorithm outperforms the PHD implementation in terms of accuracy by around 10%.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125180088","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-05-01DOI: 10.1109/SSPD.2019.8751658
Yi Li, Yang Sun, S. M. Naqvi
In recent studies, deep neural networks (DNN) have been introduced to solve monaural source separation (MSS) problem within real room environments. However, the separation performance of the existing methods is limited, especially for environments with larger RT60s. In this paper, we propose a system to train two DNNs sequentially, to mitigate the challenge and improve the separation performance. Our dereverberation mask (DM) is exploited as a training target for DNN1 and new enhanced ratio mask (ERM) is used as a training target for DNN2. The IEEE and the TIMIT corpora with real room impulse responses and noise interferences from the NOISEX dataset are used to generate speech mixtures for evaluations. The proposed method outperforms the state-of-the-art methods.
{"title":"Sequentially Trained DNNs Based Monaural Source Separation in Real Room Environments","authors":"Yi Li, Yang Sun, S. M. Naqvi","doi":"10.1109/SSPD.2019.8751658","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751658","url":null,"abstract":"In recent studies, deep neural networks (DNN) have been introduced to solve monaural source separation (MSS) problem within real room environments. However, the separation performance of the existing methods is limited, especially for environments with larger RT60s. In this paper, we propose a system to train two DNNs sequentially, to mitigate the challenge and improve the separation performance. Our dereverberation mask (DM) is exploited as a training target for DNN1 and new enhanced ratio mask (ERM) is used as a training target for DNN2. The IEEE and the TIMIT corpora with real room impulse responses and noise interferences from the NOISEX dataset are used to generate speech mixtures for evaluations. The proposed method outperforms the state-of-the-art methods.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133201725","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-05-01DOI: 10.1109/SSPD.2019.8751660
H. T. Hayvaci, Seden Hazal Gulen
In this paper, we deal with the problem of detecting point-like targets in the presence of multipath under the assumption of a partially homogeneous Gaussian disturbance with unknown covariance matrix. Therefore, we introduce an unknown scaling factor which represents the mismatch between the noise covariance matrices of test and training signals. Besides, we model the target echo as a combination of direct and multipath components where multipath echoes are thought of as scattered signals from a glistening surface which is referred to as diffuse multipath environment. Hence, the total multipath return is also represented as a Gaussian distributed random vector with an unknown covariance matrix. At the design stage, we construct a constrained Generalized Likelihood Ratio Test (GLRT) by assuming that the total primary data covariance structure, in the target present case, resembles to the covariance matrix obtained from secondary data up to a degree (related to noise scaling factor and multipath contribution). Finally, at the analysis stage, we compared the developed algorithm with the existing solutions available in the open literature. The results highlight that the new detector copes well with severe multipath conditions and has considerable scale-invariance.
{"title":"Adaptive Detection with Diffuse Multipath Exploitation in Partially Homogeneous Environments","authors":"H. T. Hayvaci, Seden Hazal Gulen","doi":"10.1109/SSPD.2019.8751660","DOIUrl":"https://doi.org/10.1109/SSPD.2019.8751660","url":null,"abstract":"In this paper, we deal with the problem of detecting point-like targets in the presence of multipath under the assumption of a partially homogeneous Gaussian disturbance with unknown covariance matrix. Therefore, we introduce an unknown scaling factor which represents the mismatch between the noise covariance matrices of test and training signals. Besides, we model the target echo as a combination of direct and multipath components where multipath echoes are thought of as scattered signals from a glistening surface which is referred to as diffuse multipath environment. Hence, the total multipath return is also represented as a Gaussian distributed random vector with an unknown covariance matrix. At the design stage, we construct a constrained Generalized Likelihood Ratio Test (GLRT) by assuming that the total primary data covariance structure, in the target present case, resembles to the covariance matrix obtained from secondary data up to a degree (related to noise scaling factor and multipath contribution). Finally, at the analysis stage, we compared the developed algorithm with the existing solutions available in the open literature. The results highlight that the new detector copes well with severe multipath conditions and has considerable scale-invariance.","PeriodicalId":281127,"journal":{"name":"2019 Sensor Signal Processing for Defence Conference (SSPD)","volume":"62 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126959282","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：