The bandwidth limitation of wideband (WB) audio systems degrades the subjective quality and naturalness of audio signals. In this paper, a new method for blind bandwidth extension of WB audio signals is proposed based on non-linear prediction and hidden Markov model (HMM). The high-frequency (HF) components in the band of 7–14 kHz are artificially restored only from the low-frequency information of the WB audio. State-space reconstruction is used to convert the fine spectrum of WB audio to a multi-dimensional space, and a non-linear prediction based on nearest-neighbor mapping is employed in the state space to restore the fine spectrum of the HF components. The spectral envelope of the resulting HF components is estimated based on an HMM according to the features extracted from the WB audio. In addition, the proposed method and the reference methods are applied to the ITU-T G.722.1 WB audio codec for comparison with the ITU-T G.722.1C super WB audio codec. Objective quality evaluation results indicate that the proposed method is preferred over the reference bandwidth extension methods. Subjective listening results show that the proposed method has a comparable audio quality with G.722.1C and improves the extension performance compared with the reference methods.

Linear finite transducers underlie a series of schemes for Public Key Cryptography (PKC)proposed in the 90s of the last century. The uninspiring and arid language then used,condemned these works to oblivion. Although some of these schemes were afterwards shown tobe insecure, the promise of a new system of PKC relying on different complexityassumptions is still quite exciting. The algorithms there used depend heavily on theresults of invertibility of linear transducers. In this paper we introduce the notion ofpost-initial linear transducer, which is an extension of the notion of linear finitetransducer with memory, and for which the previous fundamental results on invertibilitystill hold. This extension enabled us to give a new method to obtain a left inverse of anyinvertible linear finite transducer with memory. It also plays an essencial role in thenecessary and sufficient condition that we give for left invertibility of linear finitetransducers.

In this invited paper, my overview material on the same topic as presented in the plenary overview session of APSIPA-2011 and the tutorial material presented in the same conference [1] are expanded and updated to include more recent developments in deep learning. The previous and the updated materials cover both theory and applications, and analyze its future directions. The goal of this tutorial survey is to introduce the emerging area of deep learning or hierarchical learning to the APSIPA community. Deep learning refers to a class of machine learning techniques, developed largely since 2006, where many stages of non-linear information processing in hierarchical architectures are exploited for pattern classification and for feature learning. In the more recent literature, it is also connected to representation learning, which involves a hierarchy of features or concepts where higher-level concepts are defined from lower-level ones and where the same lower-level concepts help to define higher-level ones. In this tutorial survey, a brief history of deep learning research is discussed first. Then, a classificatory scheme is developed to analyze and summarize major work reported in the recent deep learning literature. Using this scheme, I provide a taxonomy-oriented survey on the existing deep architectures and algorithms in the literature, and categorize them into three classes: generative, discriminative, and hybrid. Three representative deep architectures – deep autoencoders, deep stacking networks with their generalization to the temporal domain (recurrent networks), and deep neural networks (pretrained with deep belief networks) – one in each of the three classes, are presented in more detail. Next, selected applications of deep learning are reviewed in broad areas of signal and information processing including audio/speech, image/vision, multimodality, language modeling, natural language processing, and information retrieval. Finally, future directions of deep learning are discussed and analyzed.

Consider an absolutely continuous distribution on [0, ∞) with known mean μ, and hazard rate function, h satisfying, 0<a≤h(t)≤b<∞, for almost all t≥0. We derive the sharp range for σ2, under these constraints.

Many studies that gather social network data use survey methods that lead to censored, missing, or otherwise incomplete information. For example, the popular fixed rank nomination (FRN) scheme, often used in studies of schools and businesses, asks study participants to nominate and rank at most a small number of contacts or friends, leaving the existence of other relations uncertain. However, most statistical models are formulated in terms of completely observed binary networks. Statistical analyses of FRN data with such models ignore the censored and ranked nature of the data and could potentially result in misleading statistical inference. To investigate this possibility, we compare Bayesian parameter estimates obtained from a likelihood for complete binary networks with those obtained from likelihoods that are derived from the FRN scheme, and therefore accommodate the ranked and censored nature of the data. We show analytically and via simulation that the binary likelihood can provide misleading inference, particularly for certain model parameters that relate network ties to characteristics of individuals and pairs of individuals. We also compare these different likelihoods in a data analysis of several adolescent social networks. For some of these networks, the parameter estimates from the binary and FRN likelihoods lead to different conclusions, indicating the importance of analyzing FRN data with a method that accounts for the FRN survey design.