The combination of Automatic Speech Recognition (ASR) systems generally relies on a posteriori merge of system outputs or on a cross-adaptation. In this paper, we propose an integrated approach where the search of a p...
详细信息
ISBN:
(纸本)1424407281
The combination of Automatic Speech Recognition (ASR) systems generally relies on a posteriori merge of system outputs or on a cross-adaptation. In this paper, we propose an integrated approach where the search of a primary system is driven by the outputs of a secondary one. This method allows to drive the primary system search by using the one-best hypotheses and the word posteriors gathered from the secondary system. Experiments are carried out within the experimental framework of the ESTER evaluation campaign [1]. Results show that the driven decoding algorithm significantly outperforms the two single ASR systems (-8% of relative WER, -1.7% absolute). Finally, we investigate the interactions between driven decoding and cross-adaptations. The best cross-adaptation strategy in combination with the driven decoding process brings to a final absolute gain of about 1.9% WER.
Reed Solomon codes are indeed an elegant illustration of a very practical application of abstract algebra. An interesting discovery was the fact that Euclid's algorithm for finding greatest common divisors of poly...
详细信息
Reed Solomon codes are indeed an elegant illustration of a very practical application of abstract algebra. An interesting discovery was the fact that Euclid's algorithm for finding greatest common divisors of polynomials, could be used for decoding these codes. In this article we explain this connection.
We consider the guessing secrets problem defined by Chung et al. [2001]. This is a variant of the standard 20 questions game where the player has a set of k > 1 secrets from a universe of N possible secrets. The pl...
详细信息
We consider the guessing secrets problem defined by Chung et al. [2001]. This is a variant of the standard 20 questions game where the player has a set of k > 1 secrets from a universe of N possible secrets. The player is asked Boolean questions about the secret. For each question. the player picks one of the k secrets adversarially, and answers according to this secret. We present an explicit set of O(log N) questions together with an efficient (i.e., poly(log N) time) algorithm to solve the guessing secrets problem for the case of 2 secrets. This answers the main algorithmic question left unanswered by Chung et al. [2001]. The main techniques we use are small e- biased spaces and the notion of list decoding. We also establish bounds on the number of questions needed to solve the k-secrets game for k > 2, and discuss how list decoding can be used to get partial information about the secrets, specifically to find a small core of secrets that must intersect the actual set of k secrets.
Neural spike train decoding algorithms are important tools for characterizing how ensembles of neurons represent biological signals. We present a Bayesian neural spike train decoding algorithm based on a point process...
详细信息
Neural spike train decoding algorithms are important tools for characterizing how ensembles of neurons represent biological signals. We present a Bayesian neural spike train decoding algorithm based on a point process model of individual neurons, a linear stochastic state-space model of the biological signal, and a temporal latency parameter. The latency parameter represents the temporal lead or lag between the biological signal and the ensemble spiking activity. We use the algorithm to study Whether the representation of position by the ensemble spiking activity of pyramidal neurons in the CA1 region of the rat hippocampus is more consistent with prospective coding, i.e., future position, or retrospective coding, past position. Using 44 simultaneously recorded neurons and an ensemble delay latency of 400 ms, the median decoding error was 5.1 cm during 10 min of foraging in an open circular environment. The true coverage probability for the algorithm's 0.95 confidence regions was 0.71. These results illustrate how the Bayesian neural spike train decoding paradigm may be used to investigate spatio-temporal representations of position by an ensemble of hippocampal neurons.
We propose a simple decoder for a widely used array code, known as the EVENODD code, which is originally designed to correct phased burst errors, to make it useful for correcting nonphased errors. The proposed scheme ...
详细信息
We propose a simple decoder for a widely used array code, known as the EVENODD code, which is originally designed to correct phased burst errors, to make it useful for correcting nonphased errors. The proposed scheme is capable of correcting almost all bursts up to a certain length. We show that the failure rate is sufficiently small and approaches zero as the block length increases. The redundancy of the code is twice the maximal burst length, which is a lower bound for the redundancy of a true burst-error-correcting code. Both the encoder and the decoder have very low complexity, both in terms of number of operations and in terms of computer code size.
Over the past ten years, we have tested and helped develop a multi-electrode array for chronic cortical recordings in behaving non-human primates. We have found that it is feasible to record from dozens of single unit...
详细信息
ISBN:
(纸本)0780387406
Over the past ten years, we have tested and helped develop a multi-electrode array for chronic cortical recordings in behaving non-human primates. We have found that it is feasible to record from dozens of single units in the motor cortex for extended periods of time and that these signals can be decoded in a closed-loop, real-time system to generate goal-directed behavior of external devices. This work has culminated in a FDA clinical trial that has demonstrated that a tetraplegic patient can voluntarily modulate motor cortical activity in order to move a computer cursor to visual targets. Further advances in BMI technology using non-human primates have focused on using multiple modes of control from signals in different cortical areas. We demonstrate that primary motor cortical activity may be optimized for continuous movement control whereas signals from the premotor cortex may be better suited for discrete target selection. We propose a hybrid BMI whereby decoding can be voluntarily switched from discrete to continuous control modes.
A simple technique to control transient chaos in the turbo-decoding algorithm, which improves the performance of classical turbo codes, is developed. On average, the turbo-decoding algorithm with control (when stopped...
详细信息
A simple technique to control transient chaos in the turbo-decoding algorithm, which improves the performance of classical turbo codes, is developed. On average, the turbo-decoding algorithm with control (when stopped after eight iterations) shows a gain of 0.25-0.3 dB in the practical signal-to-noise ratios (from 0 to 1 dB), compared to the classical turbo-decoding algorithm without control.
In this paper, we study a wide range of graph-based message-passing schedules for iterative decoding of low-density parity-check (LDPC) codes. Using the Tanner graph (TG) of the code and for different nodes and edges ...
详细信息
In this paper, we study a wide range of graph-based message-passing schedules for iterative decoding of low-density parity-check (LDPC) codes. Using the Tanner graph (TG) of the code and for different nodes and edges of the graph, we relate the first iteration in which the corresponding messages deviate from their optimal value (corresponding to a cycle-free graph) to the girths and the lengths of the shortest closed walks in the graph. Using this result, we propose schedules, which are designed based on the distribution of girths and closed walks in the TG of the code, and categorize them as node based versus edge based, unidirectional versus bidirectional, and deterministic versus probabilistic. These schedules, in some cases, outperform the previously known. schedules, and in other cases, provide less complex alternatives with more or less the same performance. The performance/complexity tradeoff and the best choice of schedule appear to depend not only on the girth and closed-walk distributions of the TG, but also on the iterative decoding algorithm and channel characteristics. We examine the application of schedules to belief propagation (sum-product) over additive white Gaussian noise (AWGN) and Rayleigh fading channels, min-sum (max-sum) over an AWGN channel, and Gallager's algorithm A over a binary symmetric channel.
In this letter, we show that a concatenated zigzag code can be viewed as a low-density parity-check (LDPC) code. Based on the bipartite graph representation for such a parallel-concatenated code, various sum-product b...
详细信息
In this letter, we show that a concatenated zigzag code can be viewed as a low-density parity-check (LDPC) code. Based on the bipartite graph representation for such a parallel-concatenated code, various sum-product based decoding algorithms are introduced and compared. The results show that the improved versions of sum-product algorithm exhibit better convergence rate while maintaining the essential parallel form.
Cryptography and Coding Theory are closely related in many respects. Recently, the problem of "decoding Reed Solomon codes" (also known as "polynomial reconstruction") was suggested as an intractab...
详细信息
Cryptography and Coding Theory are closely related in many respects. Recently, the problem of "decoding Reed Solomon codes" (also known as "polynomial reconstruction") was suggested as an intractability assumption to base the security of protocols on. This has initiated a line of cryptographic research exploiting the rich algebraic structure of the problem and its variants. In this paper we give a short overview of the recent works in this area as well as list directions and open problems in Polynomial Reconstruction Based Cryptography.
暂无评论