• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.14 no.4
• /
• pp.332-339
• /
• 2014

Face classification has wide applications in security and surveillance. However, this technique presents various challenges caused by pose, illumination, and expression changes. Face recognition with long-distance images involves additional challenges, owing to focusing problems and motion blurring. Multiple frames under varying spatial or temporal settings can acquire additional information, which can be used to achieve improved classification performance. This study investigates the effectiveness of multi-frame decision-level fusion with photon-counting linear discriminant analysis. Multiple frames generate multiple scores for each class. The fusion process comprises three stages: score normalization, score validation, and score combination. Candidate scores are selected during the score validation process, after the scores are normalized. The score validation process removes bad scores that can degrade the final output. The selected candidate scores are combined using one of the following fusion rules: maximum, averaging, and majority voting. Degraded facial images are employed to demonstrate the robustness of multi-frame decision-level fusion in harsh environments. Out-of-focus and motion blurring point-spread functions are applied to the test images, to simulate long-distance acquisition. Experimental results with three facial data sets indicate the efficiency of the proposed decision-level fusion scheme.

• Proceedings of the Acoustical Society of Korea Conference
• /
• autumn
• /
• pp.131-134
• /
• 2000

본 논문에서는 각 지역의 수중 음향 센서로부터 중앙의 정보 융합 센터로 전송되어진 동일한 또는 상이한 표적의 Identity 정보들을 종합해 최종적으로 표적의 Identity를 결정하는 Decision Fusion 기법을 다룬다. 기존의 연구는 표적의 속성 정보로부터 정보 융합을 통해 표적의 Identity를 선택하는 기법을 주로 다루고 있다. 그러나 본 논문에서는 기존의 연구보다 한 단계 나아가 선택된 표적의 Identity들로부터 운용자가 가장 합리적인 결정을 내릴 수 있도록 하는 표적의 Identity 결정을 위한 Decision Fusion 기법을 제안한다. 이러한 수중 음향 표적 식별 시스템에서의 Identity Decision Fusion 기법으로 Voting 기법, 센서 정보의 신뢰도를 고려한 Weighted Voting 기법, 그리고 다 기준 의사 결정 기법인 Analytic Hierarchy Process (AHP) 기법을 제안하고 그 성능을 평가한다

• International journal of advanced smart convergence
• /
• v.9 no.1
• /
• pp.193-201
• /
• 2020

The main objective of this study is to investigate the impact of additional modalities on the performance of emotion recognition using speech, facial expression and physiological measurements. In order to compare different approaches, we designed a feature-based recognition system as a benchmark which carries out linear supervised classification followed by the leave-one-out cross-validation. For the classification of four emotions, it turned out that bimodal fusion in our experiment improves recognition accuracy of unimodal approach, while the performance of trimodal fusion varies strongly depending on the individual. Furthermore, we experienced extremely high disparity between single class recognition rates, while we could not observe a best performing single modality in our experiment. Based on these observations, we developed a novel fusion method, called parametric decision fusion (PDF), which lies in building emotion-specific classifiers and exploits advantage of a parametrized decision process. By using the PDF scheme we achieved 16% improvement in accuracy of subject-dependent recognition and 10% for subject-independent recognition compared to the best unimodal results.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.5 no.4
• /
• pp.649-664
• /
• 2011

Cooperative spectrum sensing (CSS) with decision fusion is considered as a key technology for tackling the challenges caused by fading/shadowing effects and noise uncertainty in spectrum sensing in cognitive radio. However, most existing solutions assume an error-free decision transmission, which is obviously not the case in realistic scenarios. This paper extends the general decision-fusion-based CSS scheme by considering the fading/shadowing effects and noise corruption in the common control channels. With this more practical model, the fusion centre first estimates the local decisions using a binary minimum error probability detector, and then combines them to get the final result. Theoretical analysis and simulation of this CSS scheme are performed over typical channels, which suggest some performance deterioration compared with the pure case that assumes an error-free decision transmission. Furthermore, the fusion strategy optimization in the proposed cooperation model is also investigated using the Bayesian criteria. The numerical results show that the total error rate of noisy CSS is higher than that of the pure case, and the optimal values of fusion parameter in the counting rule under both cases decrease as the local detection threshold increases.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.5
• /
• pp.410-419
• /
• 2004

In this paper. we propose a method of confidence measure fusion under Bayesian framework for speech recognition. Centralized and distributed schemes are considered for confidence measure fusion. Centralized fusion is feature level fusion which combines the values of individual confidence scores and makes a final decision. In contrast. distributed fusion is decision level fusion which combines the individual decision makings made by each individual confidence measuring method. Optimal Bayesian fusion rules for centralized and distributed cases are presented. In isolated word Out-of-Vocabulary (OOV) rejection experiments. centralized Bayesian fusion shows over 13% relative equal error rate (EER) reduction compared with the individual confidence measure methods. In contrast. the distributed Bayesian fusion shows no significant performance increase.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.1999-2005
• /
• 2005

Collaboration in wireless sensor networks must be fault-tolerant due to the harsh environmental conditions in which such networks can be deployed. This paper focuses on finding signal processing algorithms for collaborative target detection based on the generalized approach to signal processing in the presence of noise that are efficient in terms of communication cost, precision, accuracy, and number of faulty sensors tolerable in the wireless sensor network. Two algorithms, namely, value fusion and decision fusion constructed according to the generalized approach to signal processing in the presence of noise, are identified first. When comparing their performance and communication overhead, decision fusion is found to become superior to value fusion as the ratio of faulty sensors to fault free sensors increases. The use of the generalized approach to signal processing in the presence of noise under designing value and decision fusion algorithms in wireless sensor networks allows us to obtain the same performance, but at low values of signal energy, as under the employment of universally adopted signal processing algorithms widely used in practice.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.20 no.5
• /
• pp.133-138
• /
• 2010

We present a new multimodal biometric authentication method, which performs both feature-level fusion and decision-level fusion. After generating support vector machines for new features made by integrating face and voice features, the final decision for authentication is made by integrating decisions of face SVM classifier, voice SVM classifier and integrated features SVM clssifier. We justify our proposal by comparing our method with traditional one by experiments with XM2VTS multimodal database. The experiments show that our multilevel fusion algorithm gives higher recognition rate than the existing schemes.

• Journal of the Institute of Convergence Signal Processing
• /
• v.12 no.2
• /
• pp.113-118
• /
• 2011

In this paper, we develop a speaker gender classification technique using collaborative sensor fusion for use in a wireless sensor network. The distributed sensor nodes remove the unwanted input data using the BER(Band Energy Ration) based voice activity detection, process only the relevant data, and transmit the hard labeled decisions to the fusion center where a global decision fusion is carried out. This takes advantages of power consumption and network resource management. The Bayesian sensor fusion and the global weighting decision fusion methods are proposed to achieve the gender classification. As the number of the sensor nodes varies, the Bayesian sensor fusion yields the best classification accuracy using the optimal operating points of the ROC(Receiver Operating Characteristic) curves_ For the weights used in the global decision fusion, the BER and MCL(Mutual Confidence Level) are employed to effectively combined at the fusion center. The simulation results show that as the number of the sensor nodes increases, the classification accuracy was even more improved in the low SNR(Signal to Noise Ration) condition.

• The Korean Journal of Applied Statistics
• /
• v.19 no.2
• /
• pp.291-303
• /
• 2006

Data mining is the work to extract information from existing data file. So, the one of best important thing in data mining process is the quality of data to be used. In this thesis, we propose the data fusion technique using decision rule for data enrichment that one phase to improve data quality in KDD process. Simulations were performed to compare the proposed data fusion technique with the existing techniques. As a result, our data fusion technique using decision rule is characterized with low MSE or misclassification rate in fusion variables.

• Proceedings of the IEEK Conference
• /
• 2006.06a
• /
• pp.97-98
• /
• 2006

Fusion of decisions from multiple distributed sensor nodes is studied in this work. Based on the canonical parallel fusion model, we derive the optimal likelihood ratio based fusion rule with the assumptions of the generalized Gaussian noise model and the arbitrary fading channel. This optimal fusion rule, however, requires the complete knowledge of the channels and the detection performance of local sensor nodes. To mitigate these requirements and to provide near optimum performance, we derive suboptimum fusion rules by using high and low signal-to-noise ratio (SNR) approximations to the optimal fusion rule. Performance evaluation is conducted through simulations.