In Natural Language Processing, measuring word frequency combined with word distribution can yield a precise indicator of lexical relevance, a measure of great value in the context of Information Retrieval. Such detection of keywords exploits the structural properties of text as revealed notably by Zipf’s Law which describes frequency distribution as a ‘long tailed’ phenomenon. Can such properties be found in images? If so, can they serve to distinguish high content items (particular colours coded as RGBs) from low information items? To explore this possibility, we have applied NLP algorithms to a corpus of satellite images in order to extract a number of linguistic-type features in bitmaps so as to augment the original corpus with distributional information regarding its RGBs and observe if this addition improves accuracy throughout a Machine Learning pipeline tested with several Transfer Learning models.
{"title":"An NLP approach to Image Analysis","authors":"G. Martínez","doi":"10.56541/kfbi5107","DOIUrl":"https://doi.org/10.56541/kfbi5107","url":null,"abstract":"In Natural Language Processing, measuring word frequency combined with word distribution can yield a precise indicator of lexical relevance, a measure of great value in the context of Information Retrieval. Such detection of keywords exploits the structural properties of text as revealed notably by Zipf’s Law which describes frequency distribution as a ‘long tailed’ phenomenon. Can such properties be found in images? If so, can they serve to distinguish high content items (particular colours coded as RGBs) from low information items? To explore this possibility, we have applied NLP algorithms to a corpus of satellite images in order to extract a number of linguistic-type features in bitmaps so as to augment the original corpus with distributional information regarding its RGBs and observe if this addition improves accuracy throughout a Machine Learning pipeline tested with several Transfer Learning models.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114978554","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}
For the implementation of Autonomously navigating Unmanned Air Vehicles (UAV) in the real world, it must be shown that safe navigation is possible in all real-world scenarios. In the case of UAVs powered by Deep Learning algorithms, this is a difficult task to achieve, as the weak point of any trained network is the reduction in predictive capacity when presented with unfamiliar input data. It is possible to train for more use cases, however more data is required for this, requiring time and manpower to acquire. In this work, a potential solution to the manpower issues of exponentially scaling dataset size and complexity is presented, through the generation of artificial image datasets that are based off of a 3D scanned recreation of a physical space and populated with 3D scanned objects of a specific class. This simulation is then used to generate image samples that iterates temporally resulting in a slice-able dataset that contains time varied components of the same class.
{"title":"Reality Analagous Synthetic Dataset Generation with Daylight Variance for Deep Learning Classification","authors":"Thomas Lee, Susan Mckeever, J. Courtney","doi":"10.56541/poya9239","DOIUrl":"https://doi.org/10.56541/poya9239","url":null,"abstract":"For the implementation of Autonomously navigating Unmanned Air Vehicles (UAV) in the real world, it must be shown that safe navigation is possible in all real-world scenarios. In the case of UAVs powered by Deep Learning algorithms, this is a difficult task to achieve, as the weak point of any trained network is the reduction in predictive capacity when presented with unfamiliar input data. It is possible to train for more use cases, however more data is required for this, requiring time and manpower to acquire. In this work, a potential solution to the manpower issues of exponentially scaling dataset size and complexity is presented, through the generation of artificial image datasets that are based off of a 3D scanned recreation of a physical space and populated with 3D scanned objects of a specific class. This simulation is then used to generate image samples that iterates temporally resulting in a slice-able dataset that contains time varied components of the same class.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124829159","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}
An approach for measuring the distance between two smartphones is presented in this paper. The method uses each smartphone’s microphone(s) and speaker(s) to concurrently emit and record audio in order to calculate the sound propagation delay and hence distance. Each device in turn emits a different audible pseudo-random binary sequence (PRBS) - specifically, a maximum length sequence (MLS). Each device captures both emitted signals in one continuous recording. The propagation delay between the devices is calculated by comparing their respective recordings, and in particular the temporal positions of the emitted signals within each recording. Each device emits one of the signals, records both signals, and then sends its recording to a master device for analysis, which is performed by a custom web application and is therefore independent of operating system. A mean error of 32.29 mm was found in initial testing, which was conducted using Samsung Galaxy A10 devices running Android 10. The key innovation in this method is that it requires no clock time synchronisation between devices because the distance is determined by comparing inter-transmission delays in the two recordings. Potential future improvements are discussed, including how to take into account the exact locations of each phone’s microphone and speaker to increase accuracy.
{"title":"Distance measurement between smartphones within an ad-hoc camera array, using audible PRBS","authors":"Pádraic McEvoy, D. Berry, Ted Burke","doi":"10.56541/kebp4512","DOIUrl":"https://doi.org/10.56541/kebp4512","url":null,"abstract":"An approach for measuring the distance between two smartphones is presented in this paper. The method uses each smartphone’s microphone(s) and speaker(s) to concurrently emit and record audio in order to calculate the sound propagation delay and hence distance. Each device in turn emits a different audible pseudo-random binary sequence (PRBS) - specifically, a maximum length sequence (MLS). Each device captures both emitted signals in one continuous recording. The propagation delay between the devices is calculated by comparing their respective recordings, and in particular the temporal positions of the emitted signals within each recording. Each device emits one of the signals, records both signals, and then sends its recording to a master device for analysis, which is performed by a custom web application and is therefore independent of operating system. A mean error of 32.29 mm was found in initial testing, which was conducted using Samsung Galaxy A10 devices running Android 10. The key innovation in this method is that it requires no clock time synchronisation between devices because the distance is determined by comparing inter-transmission delays in the two recordings. Potential future improvements are discussed, including how to take into account the exact locations of each phone’s microphone and speaker to increase accuracy.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125104420","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}
Low-resolution thermal sensing technology is suitable for sleep monitoring due to being light invariant and privacy preserving. Feature extraction is a critical step in facilitating robust detection and tracking, therefore this paper compares a blob analysis approach of extracting statistical features to several common feature descriptor algorithm approaches (SURF and KAZE). The features are extracted from thermal binary image data for the purpose of detecting bed occupancy. Four common machine learning models (SVM, KNN, DT and NB) were trained and evaluated using a leave-one-subject-out validation method. The SVM trained with feature descriptor data achieved the highest accuracy of 0.961.
{"title":"A Comparison of Feature Extraction Methods Applied to Thermal Sensor Binary Image Data to Classify Bed Occupancy","authors":"Rebecca Hand, I. Cleland, C. Nugent","doi":"10.56541/qlzv1440","DOIUrl":"https://doi.org/10.56541/qlzv1440","url":null,"abstract":"Low-resolution thermal sensing technology is suitable for sleep monitoring due to being light invariant and privacy preserving. Feature extraction is a critical step in facilitating robust detection and tracking, therefore this paper compares a blob analysis approach of extracting statistical features to several common feature descriptor algorithm approaches (SURF and KAZE). The features are extracted from thermal binary image data for the purpose of detecting bed occupancy. Four common machine learning models (SVM, KNN, DT and NB) were trained and evaluated using a leave-one-subject-out validation method. The SVM trained with feature descriptor data achieved the highest accuracy of 0.961.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133368388","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}
Wang Yao, Viktor Varkarakis, Joseph Lemley, P. Corcoran
Recent low-power neural accelerator hardware provides a solution for end-to-end privacy and secure facial authentication, such as smart refueling machine locks in shared accommodation, smart speakers, or televisions that respond only to family members. This work explores the impact that head pose variation has on the performance of a state-of-the-art face recognition model. A synthetic technique is employed to introduce head pose variation into data samples. Experiments show that the synthetic pose variations have a similar effect on face recognition performance as the real samples with pose variations. The impact of large variations of head poses on the face recognizer was then explored by further amplifying the angle of the synthetic head pose. It is found that the accuracy of the face recognition model deteriorates as the pose increases. After fine-tuning the network, the face recognition model achieves close to the accuracy of frontal faces in all pose variations, indicating that the face recognition model can be tuned to compensate for the effect of large poses.
{"title":"On the Feasibility of Privacy-Secured Facial Authentication for low-power IoT Devices - Quantifying the Effects of Head Pose Variation on End-to-End Neural Face Recognition","authors":"Wang Yao, Viktor Varkarakis, Joseph Lemley, P. Corcoran","doi":"10.56541/fevr2516","DOIUrl":"https://doi.org/10.56541/fevr2516","url":null,"abstract":"Recent low-power neural accelerator hardware provides a solution for end-to-end privacy and secure facial authentication, such as smart refueling machine locks in shared accommodation, smart speakers, or televisions that respond only to family members. This work explores the impact that head pose variation has on the performance of a state-of-the-art face recognition model. A synthetic technique is employed to introduce head pose variation into data samples. Experiments show that the synthetic pose variations have a similar effect on face recognition performance as the real samples with pose variations. The impact of large variations of head poses on the face recognizer was then explored by further amplifying the angle of the synthetic head pose. It is found that the accuracy of the face recognition model deteriorates as the pose increases. After fine-tuning the network, the face recognition model achieves close to the accuracy of frontal faces in all pose variations, indicating that the face recognition model can be tuned to compensate for the effect of large poses.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125932219","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}
Pierre Matysiak, Susana Ruano, Martin Alain, A. Smolic
Neural Radiance Fields (NeRF) is a recent technology which had a large impact in computer vision, promising to generate high quality novel views and corresponding disparity map, all using a fairly small number of input images. In effect, they are a new way to represent a light field. In this paper, we compare NeRF with traditional light field methods for novel view synthesis and depth estimation, in an attempt to quantify the advantages brought by NeRF, and to put these results in perspective with the way both paradigms are used practically. We provide qualitative and quantitative comparisons, discuss them and highlight some aspects of working with NeRF depending on the type of light field data used.
{"title":"A Comparative Study of Traditional Light Field Methods and NeRF","authors":"Pierre Matysiak, Susana Ruano, Martin Alain, A. Smolic","doi":"10.56541/iqkc6774","DOIUrl":"https://doi.org/10.56541/iqkc6774","url":null,"abstract":"Neural Radiance Fields (NeRF) is a recent technology which had a large impact in computer vision, promising to generate high quality novel views and corresponding disparity map, all using a fairly small number of input images. In effect, they are a new way to represent a light field. In this paper, we compare NeRF with traditional light field methods for novel view synthesis and depth estimation, in an attempt to quantify the advantages brought by NeRF, and to put these results in perspective with the way both paradigms are used practically. We provide qualitative and quantitative comparisons, discuss them and highlight some aspects of working with NeRF depending on the type of light field data used.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127982842","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}
Gabriel R. Palma, Ana Aquino, P. Monticelli, L. Verdade, C. Markham, Rafael Moral
Soundscape ecologists aim to study the acoustic characteristics of an area that reflects natural processes [Schafer, 1977]. These sounds can be interpreted as biological (biophony), geophysical (geophony), and human-produced (anthrophony) [Pijanowski et al., 2011]. A common task is to use sounds to identify species based on the frequency content of a given signal. This signal can be further converted into spectrograms enabling other types of analysis to automate the identification of species. Based on the promising results of deep learning methods, such as Convolution Neural Networks (CNNs) in image classification, here we propose the use of a pre-trained VGG16 CNN architecture to identify two nocturnal avian species, namely Antrostomus rufus and Megascops choliba, commonly encountered in Brazilian forests. Monitoring the abundance of these species is important to ecologists to develop conservation programmes, detect environmental disturbances and assess the impact of human action. Specialists recorded sounds in 16-bit wave files at a sampling rate of 44Hz and classified the presence of these species. With the classified wave files, we created additional classes to visualise the performance of the VGG16 CNN architecture for detecting both species. We end up with six categories containing 60 seconds of audio of species vocalisation combinations and background only sounds. We produced spectrograms using the information from each RGB channel, only one channel (grey-scale), and applied the histogram equalisation technique to the grey-scale images. A comparison of the system performance using histogram equalised images and unmodified images was made. Histogram equalisation improves the contrast, and so the visibility to the human observer. Investigating the effect of histogram equalisation on the performance of the CNN was a feature of this study. Moreover, to show the practical application of our work, we created 51 minutes of audio, which contains more noise than the presence of both species (a scenario commonly encountered in field surveys). Our results showed that the trained VGG16 CNN produced, after 8000 epochs, a training accuracy of 100% for the three approaches. The test accuracy was 80.64%, 75.26%, and 67.74% for the RGB, grey-scaled, and histogram equalised approaches. The method’s accuracy on the synthetic audio file of 51 minutes was 92.15%. This accuracy level reveals the potential of CNN architectures in automating species detection and identification by sound using passive monitoring. Our results suggest that using coloured images to represent the spectrogram better generalises the classification than grey-scale and histogram equalised images. This study might develop future avian monitoring programmes based on passive sound recording, which significantly enhances sampling size without increasing cost.
{"title":"A machine vision system for avian song classification with CNN’s","authors":"Gabriel R. Palma, Ana Aquino, P. Monticelli, L. Verdade, C. Markham, Rafael Moral","doi":"10.56541/mhzn4111","DOIUrl":"https://doi.org/10.56541/mhzn4111","url":null,"abstract":"Soundscape ecologists aim to study the acoustic characteristics of an area that reflects natural processes [Schafer, 1977]. These sounds can be interpreted as biological (biophony), geophysical (geophony), and human-produced (anthrophony) [Pijanowski et al., 2011]. A common task is to use sounds to identify species based on the frequency content of a given signal. This signal can be further converted into spectrograms enabling other types of analysis to automate the identification of species. Based on the promising results of deep learning methods, such as Convolution Neural Networks (CNNs) in image classification, here we propose the use of a pre-trained VGG16 CNN architecture to identify two nocturnal avian species, namely Antrostomus rufus and Megascops choliba, commonly encountered in Brazilian forests. Monitoring the abundance of these species is important to ecologists to develop conservation programmes, detect environmental disturbances and assess the impact of human action. Specialists recorded sounds in 16-bit wave files at a sampling rate of 44Hz and classified the presence of these species. With the classified wave files, we created additional classes to visualise the performance of the VGG16 CNN architecture for detecting both species. We end up with six categories containing 60 seconds of audio of species vocalisation combinations and background only sounds. We produced spectrograms using the information from each RGB channel, only one channel (grey-scale), and applied the histogram equalisation technique to the grey-scale images. A comparison of the system performance using histogram equalised images and unmodified images was made. Histogram equalisation improves the contrast, and so the visibility to the human observer. Investigating the effect of histogram equalisation on the performance of the CNN was a feature of this study. Moreover, to show the practical application of our work, we created 51 minutes of audio, which contains more noise than the presence of both species (a scenario commonly encountered in field surveys). Our results showed that the trained VGG16 CNN produced, after 8000 epochs, a training accuracy of 100% for the three approaches. The test accuracy was 80.64%, 75.26%, and 67.74% for the RGB, grey-scaled, and histogram equalised approaches. The method’s accuracy on the synthetic audio file of 51 minutes was 92.15%. This accuracy level reveals the potential of CNN architectures in automating species detection and identification by sound using passive monitoring. Our results suggest that using coloured images to represent the spectrogram better generalises the classification than grey-scale and histogram equalised images. This study might develop future avian monitoring programmes based on passive sound recording, which significantly enhances sampling size without increasing cost.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122676264","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}
Dylan Do Couto, J. Butterfield, A. Murphy, K. Rafferty, Joseph Coleman
3D machine vision is a growing trend in the filed of automation for Object Of Interest (OOI) interactions. This is most notable in sectors such as unorganised bin picking for manufacturing and the integration of Autonomous Guided Vehicles (AGVs) in logistics. In the literature, there is a key focus on advancing this area of research through methods of OOI recognition and isolation to simplify more established OOI analysis operations. The main constraint in current OOI isolation methods is the loss of important data and a long process duration which extends the overall run-time of 3D machine vision operations. In this paper we propose a new method of OOI isolation that utilises a combination of classical image processing techniques to reduce OOI data loss and improve run-time efficiency. Results show a high level of data retention with comparable faster run-times to previous research. This paper also hopes to present a series of run-time data points to set a standard for future process run-time comparisons.
{"title":"Detection and Isolation of 3D Objects in Unstructured Environments","authors":"Dylan Do Couto, J. Butterfield, A. Murphy, K. Rafferty, Joseph Coleman","doi":"10.56541/afmz9460","DOIUrl":"https://doi.org/10.56541/afmz9460","url":null,"abstract":"3D machine vision is a growing trend in the filed of automation for Object Of Interest (OOI) interactions. This is most notable in sectors such as unorganised bin picking for manufacturing and the integration of Autonomous Guided Vehicles (AGVs) in logistics. In the literature, there is a key focus on advancing this area of research through methods of OOI recognition and isolation to simplify more established OOI analysis operations. The main constraint in current OOI isolation methods is the loss of important data and a long process duration which extends the overall run-time of 3D machine vision operations. In this paper we propose a new method of OOI isolation that utilises a combination of classical image processing techniques to reduce OOI data loss and improve run-time efficiency. Results show a high level of data retention with comparable faster run-times to previous research. This paper also hopes to present a series of run-time data points to set a standard for future process run-time comparisons.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131809641","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}
Much research has been carried out into the automatic restoration of archival images. This research ranges from colourisation, to damage restoration, and super-resolution. Conversely, video restoration hasremained largely unexplored. Most efforts to date have involved extending a concept from image restoration to video, in a frame-by-frame manner. These methods result in poor temporal consistency between frames. This manifests itself as temporal instability or flicker. The purpose of this work is to improve upon this limitation. This improvement will be achieved by employing a hybrid approach of deep-learning and exemplar based colourisation. Thus, informing current frame colourisation about its neighbouring frame’s colourisations and therefore alleviating the inter-frame discrepancy issues. This paper has two main contributions. Firstly, a novel end-to-end automatic video colourisation technique with enhanced flicker reduction capabilities is proposed. Secondly, six automatic exemplar acquisition algorithms are compared. The combination of these algorithms and techniques allow for an 8.5% increase in non-referenced image quality over the previous state of the art.
{"title":"Towards Temporal Stability in Automatic Video Colourisation","authors":"Rory Ward, J. Breslin","doi":"10.56541/zvhf9195","DOIUrl":"https://doi.org/10.56541/zvhf9195","url":null,"abstract":"Much research has been carried out into the automatic restoration of archival images. This research ranges from colourisation, to damage restoration, and super-resolution. Conversely, video restoration hasremained largely unexplored. Most efforts to date have involved extending a concept from image restoration to video, in a frame-by-frame manner. These methods result in poor temporal consistency between frames. This manifests itself as temporal instability or flicker. The purpose of this work is to improve upon this limitation. This improvement will be achieved by employing a hybrid approach of deep-learning and exemplar based colourisation. Thus, informing current frame colourisation about its neighbouring frame’s colourisations and therefore alleviating the inter-frame discrepancy issues. This paper has two main contributions. Firstly, a novel end-to-end automatic video colourisation technique with enhanced flicker reduction capabilities is proposed. Secondly, six automatic exemplar acquisition algorithms are compared. The combination of these algorithms and techniques allow for an 8.5% increase in non-referenced image quality over the previous state of the art.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116778621","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}
The short paper extends an acoustic signal delay estimation method to general anechoic scenario using image processing techniques. The technique proposed in this paper localizes acoustic speech sources by creating a matrix of phase versus frequency histograms, where the same phases are stacked in appropriate bins. With larger delays and multiple sources coexisting in the same matrix, it becomes cluttered with activated bins. This results in high intensity spots on the spectrogram, making source discrimination difficult. In this paper, we have employed morphological filtering, chain-coding and straight line approximations to ignore noise and enhance the target signal features. Lastly, Hough transform is used for the source localization. The resulting estimates are accurate and invariant to the sampling-rate and shall have application in acoustic source separation.
{"title":"Acoustic Source Localization Using Straight Line Approximations","authors":"Swarnadeep Bagchi, Ruairí de Fréin","doi":"10.56541/ljrb7078","DOIUrl":"https://doi.org/10.56541/ljrb7078","url":null,"abstract":"The short paper extends an acoustic signal delay estimation method to general anechoic scenario using image processing techniques. The technique proposed in this paper localizes acoustic speech sources by creating a matrix of phase versus frequency histograms, where the same phases are stacked in appropriate bins. With larger delays and multiple sources coexisting in the same matrix, it becomes cluttered with activated bins. This results in high intensity spots on the spectrogram, making source discrimination difficult. In this paper, we have employed morphological filtering, chain-coding and straight line approximations to ignore noise and enhance the target signal features. Lastly, Hough transform is used for the source localization. The resulting estimates are accurate and invariant to the sampling-rate and shall have application in acoustic source separation.","PeriodicalId":180076,"journal":{"name":"24th Irish Machine Vision and Image Processing Conference","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115383043","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：