2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)最新文献

{"title":"Detecting Perceived Appropriateness of a Robot's Social Positioning Behavior from Non-Verbal Cues","authors":"J. Vroon, G. Englebienne, V. Evers","doi":"10.1109/CogMI48466.2019.00039","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00039","url":null,"abstract":"What if a robot could detect when you think it got too close to you during its approach? This would allow it to correct or compensate for its social 'mistake'. It would also allow for a responsive approach, where that robot would reactively find suitable approach behavior through and during the interaction. We investigated if it is possible to automatically detect such social feedback cues in the context of a robot approaching a person. We collected a dataset in which our robot would repeatedly approach people (n=30) to verbally deliver a message. Approach distance and environmental noise were manipulated, and our participants were tracked (position and orientation of upper body and head). We evaluated their perception of the robot's behavior through questionnaires and found no single or joint effects of the manipulations. This showed that, in this case, personal differences are more important than contextual cues – thus highlighting the importance of responding to behavioral feedback. This dataset is being made publicly available as part of this publication † . On this dataset, we then trained a random forest classifier to infer people's perception of the robot's approach behavior from features generated from the response behaviors. This resulted in a set of relevant features that perform significantly better than chance for a participant-dependent classifier; which implies that the behaviors of our participants, even with our relatively limited tracking, contain interpretable information about their perception of the robot's behavior. Our findings demonstrate, for this specific context, that the observable behavior of people does indeed contain usable information about their subjective perception of a robot's behavior. As such they, together with the dataset, provide a stepping stone for future research into the automatic detection of such social feedback cues, e.g. with other or more fine-grained observations of people's behavior (such as facial expressions), with more sophisticated machine learning techniques, and/or in different contexts.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127455603","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}

{"title":"Self-Adaptive Tuning for Speech Enhancement Algorithm Based on Evolutionary Approach","authors":"Ryan LeBlanc, S. Selouani","doi":"10.1109/CogMI48466.2019.00012","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00012","url":null,"abstract":"In this paper a novel approach to speech enhancement algorithm tuning is presented. A self-adaptive tuning method was developed using an evolutionary optimization algorithm. Through this new approach, an improvement on the multiband spectral subtraction method was obtained by using a genetic algorithm to adaptively tune the algorithm's parameters for the speech being processed. Experimental tests using objective quality and intelligibility measures showed that the proposed artificial intelligence-based method offers better noise reduction with less signal distortion when compared to existing speech enhancement methods.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"313 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129151016","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}

{"title":"Acute Mental Stress Measurement using Brain-IoT System","authors":"Bhagyashree Shirke, Jonathan Wong, K. George","doi":"10.1109/CogMI48466.2019.00019","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00019","url":null,"abstract":"Every individual experience stress of varied intensity while performing daily tasks. Excessive stress can be harmful to human health. Hence, stress assessment is essential in preventing detrimental long-term effects. In this study, we investigate the feasibility of EEG for the measurement of acute mental stress. Also, the post-stress analysis with four distinct cases on stress-induced subjects is carried out. The experiments accomplished were conducted using an EEG headset, ThingSpeak database, and a mobile application. The subject is required to play a mobile game, which induces stress as the game level progresses. This raw EEG data is pre-processed and analyzed in MATLAB and sent over to the ThingSpeak database. When the acute level of stress is detected, the individual gets notified by the mobile application to prompt soothing music and closing eyes. Overall, the experiment concluded with reduced stress levels of the subject after closing eyes with and without music.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"77 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122630942","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}

{"title":"Automating Deep Neural Network Model Selection for Edge Inference","authors":"Bingqian Lu, Jianyi Yang, L. Chen, Shaolei Ren","doi":"10.1109/CogMI48466.2019.00035","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00035","url":null,"abstract":"The ever increasing size of deep neural network (DNN) models once implied that they were only limited to cloud data centers for runtime inference. Nonetheless, the recent plethora of DNN model compression techniques have successfully overcome this limit, turning into a reality that DNN-based inference can be run on numerous resource-constrained edge devices including mobile phones, drones, robots, medical devices, wearables, Internet of Things devices, among many others. Naturally, edge devices are highly heterogeneous in terms of hardware specification and usage scenarios. On the other hand, compressed DNN models are so diverse that they exhibit different tradeoffs in a multi-dimension space, and not a single model can achieve optimality in terms of all important metrics such as accuracy, latency and energy consumption. Consequently, how to automatically select a compressed DNN model for an edge device to run inference with optimal quality of experience (QoE) arises as a new challenge. The state-of-the-art approaches either choose a common model for all/most devices, which is optimal for a small fraction of edge devices at best, or apply device-specific DNN model compression, which is not scalable. In this paper, by leveraging the predictive power of machine learning and keeping end users in the loop, we envision an automated device-level DNN model selection engine for QoE-optimal edge inference. To concretize our vision, we formulate the DNN model selection problem into a contextual multi-armed bandit framework, where features of edge devices and DNN models are contexts and pre-trained DNN models are arms selected online based on the history of actions and users' QoE feedback. We develop an efficient online learning algorithm to balance exploration and exploitation. Our preliminary simulation results validate our algorithm and highlight the potential of machine learning for automating DNN model selection to achieve QoE-optimal edge inference.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"51 9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115344828","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}

{"title":"Social Edge Intelligence: Integrating Human and Artificial Intelligence at the Edge","authors":"Dong Wang, D. Zhang, Yang Zhang, Md. Tahmid Rashid, Lanyu Shang, Na Wei","doi":"10.1109/CogMI48466.2019.00036","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00036","url":null,"abstract":"In this vision paper, we propose a new concept, \"Social Edge Intelligence (SEI)\", where the artificial intelligence (AI) and human intelligence (HI) are tightly integrated to address a set of critical research challenges in edge computing applications. The SEI concept is motivated by two technical trends: 1) the recent rapid advancement of AI techniques in many edge and mobile applications (e.g., mobile sensing, smart homes, intelligent transportation systems), and 2) the emergence of the crowdsourcing platforms (e.g., Amazon MTurk, Waze) that are used to explore the wisdom of common individuals. We envision that an unprecedented opportunity has been unleashed to integrate AI with human intelligence at the edge of the network to obtain the best of both worlds. In this vision paper, we will review several real-world applications under the SEI vision and discuss the fundamental research challenges in implementing SEI in those applications. We also observe that approaches originated from multiple disciplines (e.g., information theory, machine learning, AI, statistics) are desirable to address the emerging challenges in SEI applications. Finally, we conclude the paper by outlining a few research directions for future work in this exciting direction.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132095902","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}

{"title":"A2Text-Net: A Novel Deep Neural Network for Sarcasm Detection","authors":"Liyuan Liu, J. Priestley, Yiyun Zhou, H. Ray, Meng Han","doi":"10.1109/CogMI48466.2019.00025","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00025","url":null,"abstract":"Sarcasm is a common form of irony in which users usually express their negative attitudes using contrary words. Predicting sarcasm is an essential part of investigating human social interaction. Improvements in classifying sarcasm have the potential to improve other dimensions of human sentiment (e.g., brand preference, political views). In face-to-face communication, the changing of voice, eye contact, physical position, etc. provides the audience with cues to detect sarcasm. However, detecting sarcasm exclusively with text is particularly challenging, given the lack of these subtle human-centric cues. In this study, we employed a new deep neural network: A2Text-Net to mimic the face-to-face speech, which integrates auxiliary variables such as punctuations, part of speech (POS), numeral, emoji, etc. to increase classification performance. The experiment results provide evidence that our A2Text-Net approach improves classification performance over conventional machine learning and deep learning algorithms.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133931216","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}

{"title":"Classification of Space Objects Using Machine Learning Methods","authors":"M. Khalil, E. Fantino, P. Liatsis","doi":"10.1109/CogMI48466.2019.00021","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00021","url":null,"abstract":"In the last decade, the number of space object has skyrocketed. Collecting and analyzing data about these objects is essential in maintaining security of space assets. Classifying unknown objects into satellites, rocket bodies and debris represents a significant milestone in the analysis process. In this context, we investigate the effectiveness of several machine learning methods in classifying real-world light curves of space objects. The light curves are represented with a set of features extracted using the feets (feATURE eXTRACTOR FOR tIME sERIES) public tool. To address the problem of class imbalance, the synthetic minority over-sampling technique (SMOTE) is applied. We also investigate the use of Principal Component Analysis (PCA) in reducing the dimensionality of the feature space, prior to classification. In the case of the original feature set, the top performing classifier is the feedforward neural network with an accuracy of 73.6%. When SMOTE is used, an improvement in accuracy of approximately 15% is observed, with the use of SVM. However, PCA-based feature transformation leads to a slight degradation in performance of around 3%, in the case of the original feature representation, and a considerable degradation of 10%-30%, when SMOTE is used.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130291245","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}

{"title":"Random Sampling Deep Learning Mechanism for Discovering Unique Property of No Specific Local Feature Images","authors":"Zonyin Shae, Zhi-Ren Tsai, Chi-Yu Chang, J. Tsai","doi":"10.1109/CogMI48466.2019.00040","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00040","url":null,"abstract":"A unique and innovative deep learning mechanism is devised and investigated to discover the underline global property of an image or group of images of no any specific local features. The images of tea, rice, and coffee, etc., are some examples. Further extend and generalize the concept, this paper makes a conjecture that every image or a group of images has its own unique global property which can be used as an ID of that image or group of images. As such, if some properties of the physical product and therefore the property in its corresponding image are altered, one should be able to detect it. This paper makes the first research attempt to address and investigate this issue. Some initial experiments by random sampling deep learning algorithm are devised to explore this conjecture. Based on the devised mechanism, a real time tea authentication application can be built allowing farmers to trustfully establish their own product brand name which is pervasively promoted by enabling customers' real time tea authentication during the purchasing. This real time authentication can be friendly achieved simply by extracting the unique global tea image property captured by customer's mobile phone.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"88 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124691984","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}

{"title":"#selfharm on Instagram: Quantitative Analysis and Classification of Non-Suicidal Self-Injury","authors":"Lei Xian, S. Vickers, Amanda L. Giordano, Jaewoo Lee, I. Kim, Lakshmish Ramaswamy","doi":"10.1109/CogMI48466.2019.00017","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00017","url":null,"abstract":"Non-Suicidal Self-Injury (NSSI) is the intentional destruction of body tissue without the intent to die. NSSI is particularly prevalent among adolescents and young adults as a means of emotional regulation. With the proliferation of social media, NSSI content is frequently being posted, viewed, and shared on popular social media platforms, which may increase social contagion among adolescents. To address this problem, this work first quantifies the prevalence of NSSI content on social media. We develop a content crawler that searches for posts, images, and videos with NSSI-related hashtags (e.g., #selfharm), downloads NSSI content from target social media platforms, and stores them in cloud storage. We then perform a trend analysis, which confirms a steep increase in NSSI posts on social media. Moreover, this work develops an image classifier to identify NSSI or non-NSSI images from social media content. Our classifier is based on the idea of weakly supervised object localization. We evaluate our NSSI classifier with more than 30K labeled NSSI images collected from social media. In our evaluation, our classifier accurately identifies NSSI images with 94% accuracy, and it outperforms state-of-the-art pre-trained models. An accurate NSSI image classifier is an essential first step to enable us and/or social media providers to protect adolescents and young adults from social contagion due to NSSI exposure through such actions as legitimate filtering mechanisms.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"71 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126397426","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}

{"title":"Toward Autonomy: Symbiotic Formal and Statistical Machine Reasoning","authors":"J. S. Mertoguno","doi":"10.1109/CogMI48466.2019.00038","DOIUrl":"https://doi.org/10.1109/CogMI48466.2019.00038","url":null,"abstract":"Different types of machine learning, statistical types, where its knowledge in contained in set of numbers, and formal types, where its knowledge is contained in set of rules or statements, have their own strengths and weaknesses. We argue that their strengths and weaknesses are complementary, and develop a concept called Learn2Reason to harness their collective strength, without inheriting their weaknesses. The efficacy of Learn2Reason concept has been successfully demonstrated in software/binary analysis and cyber security areas. Adoption of the concept significantly improve the performance and scalability of software/binary analysis and cyber security applications and tools.","PeriodicalId":116160,"journal":{"name":"2019 IEEE First International Conference on Cognitive Machine Intelligence (CogMI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129599440","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}