Search

Few-shot image classification remains a persistent challenge due to the intrinsic difficulty faced by visual recognition models in achieving generalization with limited training data. Existing methods primarily focus on exploiting marginal distributions and overlook the disparity between the product of marginals and the joint characteristic functions. This can lead to less robust feature representations. In this paper, we introduce DBDC-SSL, a method that aims to improve few-shot visual recognition models by learning a feature extractor that produces image representations that are more robust. To improve the robustness of the model, we integrate DeepBDC (DBDC) during the training process to learn better feature embeddings by effectively computing the disparity between product of the marginals and joint characteristic functions of the features. To reduce overfitting and improve the generalization of the model, we utilize an auxiliary rotation loss for self-supervised learning (SSL) in the training of the feature extractor. The auxiliary rotation loss is derived from a pretext task, where input images undergo rotation by predefined angles, and the model classifies the rotation angle based on the features it generates. Experimental results demonstrate that DBDC-SSL is able to outperform current state-of-the-art methods on 4 common few-shot image classification benchmark, which are miniImageNet, tieredImageNet, CUB and CIFAR-FS. For 5-way 1-shot and 5-way 5-shot tasks respectively, the proposed DBDC-SSL achieved the accuracy of 68.64±0.43 and 86.02±0.28 on miniImageNet, 73.88±0.48 and 89.03±0.29 on tieredImageNet, 84.67±0.39 and 94.76±0.16 on CUB, and 75.60±0.44 and 88.49±0.31 on CIFAR-FS.

Few-shot learning continues to pose a challenge as it is inherently difficult for visual recognition models to generalize with limited labeled examples. When the training data is limited, the process of training and fine-tuning the model will be unstable and inefficient due to overfitting. In this paper, we introduce NegCosIC: Negative Cosine Similarity-Invariance-Covariance Regularization, a method that aims to improve the mean accuracy from the perspective of stabilizing the fine-tuning process and regularizing variance. NegCosIC incorporates a negative simple cosine similarity loss to stabilize the parameters of the feature extractor during fine-tuning. In addition, NegCosIC integrates invariance loss and covariance loss to regularize the embeddings in order to reduce overfitting. Experimental results demonstrate that NegCosIC is able to bring substantial improvements over the current state-of-the-art methods. An in-depth worse case analysis is also conducted and shows that NegCosIC is able to outperform state-of-the-art methods on worst case accuracy. The proposed NegCosIC achieved 2.15% and 2.13% higher accuracy on miniImageNet 1-shot and 5-shot tasks, 3.22% and 2.67% higher accuracy on CUB 1-shot and 5-shot tasks, and 2.13% and 7.74% higher accuracy on CIFAR-FS 1-shot and 5-shot tasks in terms of worst-case accuracies.

This work presents a robust speaker's location detection algorithm using a single linear microphone array that is capable of detecting multiple speech sources under the assumption that there exist nonoverlapped speech segments among sources. Namely, the overlapped speech segments are treated as uncertainty and are not used for detection. The location detection algorithm is derived from a previous work (2006), where Gaussian mixture models (GMMs) are used to model location-dependent and content and speaker-independent phase difference distributions. The proposed algorithm is proven to be robust against the complex vehicular acoustics including noise, reverberation, near-filed, far-field, line-of-sight, and non-line-of-sight conditions, and microphones' mismatch. An adaptive system architecture is developed to adjust the Gaussian mixture (GM) location model to environmental noises. To deal with unmodeled speech sources as well as overlapped speech signals, a threshold adaptation scheme is proposed in this work. Experimental results demonstrate high detection accuracy in a noisy vehicular environment.

This paper discusses the relationships between the nonstationarity of sound sources and the distribution patterns of interaural phase differences (IPDs) and interaural level differences (ILDs) based on short-term frequency analysis. The amplitude variation of nonstationary sound sources is modeled by the. exponent of polynomials from the concept of moving pole model. According to the model, the sufficient condition for utilizing the distribution patterns of IPDs and ILDs to localize a nonstationary sound source is suggested and the phenomena of multiple peaks in the distribution pattern can be explained. Simulation is performed to interpret the relation between the distribution patterns of IPD and ILD and the nonstationary sound source. Furthermore, a Gaussian-mixture binaural room distribution model (GMBRDM) is proposed to model distribution patterns of IPDs and ILDs for nonstationary sound source location classification. The effectiveness and performance of the proposed GMBRDM are demonstrated by experimental results. [ABSTRACT FROM AUTHOR]

This work proposes a novel Gaussian mixture-sound field landmark model for localization applications, based on the principle that sound fields produced by sources at different locations can be distinguished in terms of their statistical patterns. The experimental results indicate that two microphones are sufficient to differentiate among the patterns. The proposed method is robust against environmental noise and performs accurately in a complex environment. Moreover, it cannot only detect the non-line-of-sight locations when the direct path between the microphones and the location is blocked, but also can distinguish the locations aligned with respect to the line connecting the microphones. However, using only two microphones, these scenarios are difficult to handle by traditional direction-of-arrival or beamforming methods in microphone array research. The experiments were conducted on a quadruped robot platform with an eRobot agent using embedded Ethernet technology. Because of its high accuracy and low-cost, this method is suitable for robot localization in real environments. The experimental results also show that the proposed method with only two microphones outperforms the conventional multiple signal classification method (MUSIC) technique with six microphones at various signal-to-noise ratios. [ABSTRACT FROM AUTHOR]

Portfolio value-at-risk (PVAR) is widely used in practice, but recent criticisms have focused on risks arising from biased PVAR estimates due to model specification errors and other problems. The PVAR estimation method proposed in this article combines generalized Pareto distribution tails with the empirical density function to model the marginal distributions for each asset in the portfolio, and a copula model is used to form a joint distribution from the fitted marginals. The copula–mixed distribution (CMX) approach converges in probability to the true marginal return distribution but is based on weaker assumptions that may be appropriate for the returns data found in practice. CMX is used to estimate the joint distribution of log returns for the Taiwan Stock Exchange (TSE) index and the associated futures contracts on SGX and TAIFEX. The PVAR estimates for various hedge portfolios are computed from the fitted CMX model, and backtesting diagnostics indicate that CMX outperforms the alternative PVAR estimators. © 2006 Wiley Periodicals, Inc. Jrl Fut Mark 26:997–1018, 2006 [ABSTRACT FROM AUTHOR]

This work proposes and implements a novel and robust adaptive speech enhancement system, which contains both time domain and frequency domain beamformers using H∞ filtering approach to provide a clean and undisturbed speech waveform and improve the speech recognition rate in vehicle environments. A microphone array data acquisition hardware is also designed and implemented for the proposed speech enhancement system. Mutually matched microphones are needed for traditional multidimensional noise reduction methods, but this requirement is not practical for consumer applications from the cost standpoint. To overcome this issue, the proposed system adapts the mismatch dynamics to maintain the theoretical performance allowing unmatched microphones to be used in an array. Furthermore, to achieve a high speech recognition performance, the speech recognizer is usually required to be retrained for different vehicle environments due to different noise characteristics and channel effects. The proposed system using the H∞ filtering approach, which makes no assumptions about noise and disturbance, is robust to the modeling error in a channel recovery process. Consequently, the real vehicular experimental results show that the proposed frequency domain beamformer provides a satisfactory speech recognition performance without the need to retrain the speech recognizer. [ABSTRACT FROM AUTHOR]

Human-computer interaction (HCI) using speech communication is becoming increasingly important, especially in driving where safety is the primary concern. Knowing the speaker's location (i.e., speaker localization) not only improves the enhancement results of a corrupted signal, but also provides assistance to speaker identification. Since conventional speech localization algorithms suffer from the uncertainties of environmental complexity and noise, as well as from the microphone mismatch problem, they are frequently not robust in practice. Without a high reliability, the acceptance of speech-based HCI would never be realized. This work presents a novel speaker's location detection method and demonstrates high accuracy within a vehicle cabinet using a single linear microphone array. The proposed approach utilize Gaussian mixture models (GMM) to model the distributions of the phase differences among the microphones caused by the complex characteristic of room acoustic and microphone mismatch. The model can be applied both in near-field and far-field situations in a noisy environment. The individual Gaussian component of a GMM represents some general location-dependent but content and speaker-independent phase difference distributions. Moreover, the scheme performs well not only in nonline-of-sight cases, but also when the speakers are aligned toward the microphone array but at difference distances from it. This strong performance can be achieved by exploiting the fact that the phase difference distributions at different locations are distinguishable in the environment of a car. The experimental results also show that the proposed method outperforms the conventional multiple signal classification method (MUSIC) technique at various SNRs. [ABSTRACT FROM AUTHOR]