Nonlinear adaptive prediction of nonstationary signals

We describe a computationally efficient scheme for the nonlinear adaptive prediction of nonstationary signals whose generation is governed by a nonlinear dynamical mechanism. The complete predictor consists of two subsections. One performs a nonlinear mapping from the input space to an intermediate space with the aim of linearizing the input signal, and the other performs a linear mapping from the new space to the output space. The nonlinear subsection consists of a pipelined recurrent neural network (PRNN), and the linear section consists of a conventional tapped-delay-line (TDL) filter. The nonlinear adaptive predictor described is of general application. The dynamic behavior of the predictor is demonstrated for the case of a speech signal; for this application, it is shown that the nonlinear adaptive predictor outperforms the traditional linear adaptive scheme in a significant way     

Robust backward adaptive formant prediction for speech coder

To improve the error performance of speech coders, an adaptation method for the backward adapted formant predictor is proposed. The filtered residual signal is used instead of the reconstructed output signal as the input to an adaptation of the formant predictor. The performance of the filtered-residual driven adaption method in the noise free channel is as good as that of conventional output driven adaptation. Moreover, the new adaptation method maintains the same robustness to channel errors as residual driven adaptation     

A posteriori real-time recurrent learning schemes for a recurrentneural network based nonlinear predictor

Recurrent neural networks (RNNs) are well established for the nonlinear and nonstationary signal prediction paradigm. Appropriate learning algorithms, such as the real-time recurrent learning (RTRL) algorithm, have been developed for that purpose. However, little is known about the RNN time-management policy. Here, insight is provided into the time-management of the RNN, and an a posteriori approach to the RNN based nonlinear signal prediction paradigm is offered. Based upon the chosen time-management policy, algorithms are developed, from the a priori learning-a priori error strategy through to the a posteriori learning-a posteriori error strategy. Compared with the a priori algorithms, the a posteriori algorithms offered are shown to provide a better prediction performance with little further expense in terms of computational complexity. Simulations undertaken on speech using the newly introduced algorithms confirm the theoretical results     

Nonlinear adaptive prediction of complex-valued signals by complex-valued PRNN

A complex-valued pipelined recurrent neural network (CPRNN) for nonlinear adaptive prediction of complex nonlinear and nonstationary signals is introduced. This architecture represents an extension of the recently proposed real-valued PRNN of Haykin and Li in 1995. To train the CPRNN, a complex-valued real time recurrent learning (CRTRL) algorithm is first derived for a single recurrent neural network (RNN). This algorithm is shown to be generic and applicable to general signals that have complex domain representations. The CRTRL is then extended to suit the modularity of the CPRNN architecture. Further, to cater to the possibly large dynamics of the input signals, a gradient adaptive amplitude of the nonlinearity within the neurons is introduced to give the adaptive amplitude CRTRL (AACRTRL). A comprehensive analysis of the architecture and associated learning algorithms is undertaken, including the role of the number of nested modules, number of neurons within the modules, and input memory of the CPRNN. Simulations on real-world and synthetic complex data support the proposed architecture and algorithms.     

Optimal nonlinear adaptive prediction and modeling of MPEG video inATM networks using pipelined recurrent neural networks

This paper investigates the application of a pipelined recurrent neural network (PRNN) to the adaptive traffic prediction of MPEG video signal via dynamic ATM networks. The traffic signal of each picture type (I, P, and B) of MPEG video is characterized by a general nonlinear autoregressive moving average (NARMA) process. Moreover, a minimum mean-squared error predictor based on the NARMA model is developed to provide the best prediction for the video traffic signal. However, the explicit functional expression of the best mean-squared error predictor is actually unknown. To tackle this difficulty, a PRNN that consists of a number of simpler small-scale recurrent neural network (RNN) modules with less computational complexity is conducted to introduce the best nonlinear approximation capability into the minimum mean-squared error predictor model in order to accurately predict the future behavior of MPEG video traffic in a relatively short time period based on adaptive learning for each module from previous measurement data, in order to provide faster and more accurate control action to avoid the effects of excessive load situation. Since those modules of PRNN can be performed simultaneously in a pipelined parallelism fashion, this would lead to a significant improvement in the total computational efficiency of PRNN. In order to further improve the convergence performance of the adaptive algorithm for PRNN, a learning-rate annealing schedule is proposed to accelerate the adaptive learning process. Another advantage of the PRNN-based predictor is its generalization from learning that is useful for learning a dynamic environment for MPEG video traffic prediction in ATM networks where observations may be incomplete, delayed, or partially available. The PRNN-based predictor presented in this paper is shown to be promising and practically feasible in obtaining the best adaptive prediction of real-time MPEG video traffic     

An adaptive network traffic prediction approach for LDoS attacks detection

Low‐rate denial of service (LDoS) attacks reduce throughput and degrade quality of service (QoS) of network services by sending out attack packets with relatively low average rate. LDoS attack flows are difficult to detect from normal traffic since it has the property of low average rate. The research on network traffic analysis and modeling shows that network traffic measurement data are irregular nonlinear time series. To characterize and analyze network traffic between attack and non‐attack situations, the adaptive normal and abnormal ν‐support vector regression (ν‐SVR) prediction models are constructed on the basis of the reconstructed phase space. In this paper, the dimension of reconstructed phase space for ν‐SVR is optimized by Bayesian information criteria method, and the parameter in the radial basis function is adaptively adjusted by minimizing the within‐class distance and maximizing the between‐class distance in the feature space. The nonthreshold decision function is obtained through calculating the prediction error of adaptive normal and abnormal ν‐SVR prediction models, which is adopted to detect LDoS attacks. Experiments in NS‐2 environment show that the adaptive ν‐SVR prediction model can effectively predict the network traffic measurement time series, and the probability distribution of time series generated by the adaptive ν‐SVR prediction model is quite similar to that of the network traffic measurement data. Experiments also clearly demonstrate the superiority of the proposed approach in LDoS attacks detection.     

Virtex FPGA implementation of a pipelined adaptive LMS predictor for electronic support measures receivers

High-speed field-programmable gate array (FPGA) implementations of an adaptive least mean square (LMS) filter with application in an electronic support measures (ESM) digital receiver, are presented. They employ "fine-grained" pipelining, i.e., pipelining within the processor and result in an increased output latency when used in the LMS recursive system. Therefore, the major challenge is to maintain a low latency output whilst increasing the pipeline stage in the filter for higher speeds. Using the delayed LMS (DLMS) algorithm, fine-grained pipelined FPGA implementations using both the direct form (DF) and the transposed form (TF) are considered and compared. It is shown that the direct form LMS filter utilizes the FPGA resources more efficiently thereby allowing a 120 MHz sampling rate.     

Least mean square adaptive digital background calibration of pipelined analog-to-digital converters

We present an adaptive digital technique to calibrate pipelined analog-to-digital converters (ADCs). Rather than achieving linearity by adjustment of analog component values, the new approach infers component errors from conversion results and applies digital postprocessing to correct those results. The scheme proposed here draws close analogy to the channel equalization problem commonly encountered in digital communications. We show that, with the help of a slow but accurate ADC, the proposed code-domain adaptive finite-impulse-response filter is sufficient to remove the effect of component errors including capacitor mismatch, finite op-amp gain, op-amp offset, and sampling-switch-induced offset, provided they are not signal-dependent. The algorithm is all digital, fully adaptive, data-driven, and operates in the background. Strong tradeoffs between accuracy and speed of pipelined ADCs are greatly relaxed in this approach with the aid of digital correction techniques. Analog precision problems are translated into the complexity of digital signal-processing circuits, allowing this approach to benefit from CMOS device scaling in contrast to most conventional correction techniques.     

The design of a hybrid adaptive quantizer for speech codingapplications

A new adaptive quantizer which uses a combination of instantaneous and syllabic adaptation is presented for use in speech codecs. It can be designed to adapt to changes in the mean, variance, and pdf shape of its input signal, and to quantize the signal using one or more bits/sample. It is therefore called the generalized hybrid adaptive quantizer (GHAQ). An efficient procedure for optimizing the GHAQ using a training sequence of signal samples is described, and the effects on the performance of the GHAQ of varying the memory length and the syllabic compandor time constant are investigated. It is found that an optimized version of the two-bit GHAQ offers improved signal-to-noise ratio over Jayant's adaptive quantizer with a one-word memory when it is used in a predictive speech codec with a zero-, first-, or second-order fixed predictor     

Enhanced pipelined architecture of H.264/AVC intra prediction

This paper presents a high-performance encoder for H.264/AVC intra prediction. Due to long data dependency loop of intra 4×4 prediction and complex algorithms, improving encoding speed turns into a stumbling block we have to face. To solve this problem, we first propose a pipelined method in and between macro blocks with new block processing order to accelerate the encoding speed. Benefiting from the pipelined method, reconstructed pixels of up-right blocks are available for two blocks in a macro block which could not take advantage of reconstructed pixels of up-right blocks in JM. So diagonal down left mode and vertical left mode are effective for these two blocks, which ultimately achieves a better bit-rate. Secondly, all 4×4 mode formula sharing method is proposed to reduce the redundancy of predicting formulas. Thirdly, streamlined reconstruction method is applied to improve the performance of reconstruction. CAVLC encoder with three parallel units is proposed to improve entropy coding speed significantly. As a result, it takes 268 cycles to encode a macro block. The experimental results indicate that synthesized into a 0.18 µm CMOS cell library, the new architecture only requires about 238K gates and it is able to encode 1080pHD video sequences at 30 frames per second (fps), at the operating frequency of 56 MHz.     

Using adaptive linear prediction to support real-time VBR videounder RCBR network service model

A dynamic bandwidth allocation strategy to support variable bit rate (VBR) video traffic is proposed. This strategy predicts the bandwidth requirements for future frames using adaptive linear prediction that minimizes the mean square error. The adaptive technique does not require any prior knowledge of the traffic statistics nor assume stationarity. Analyses using six one-half-hour video tracts indicate that prediction errors for the bandwidth required for the next frames and group of pictures (GOP) are almost white noise or short memory. The performance of the strategy is studied using renegotiated constant bit rate (RCBR) network service model and methods that control the tradeoff between the number of renegotiations and network utilization are proposed. Simulation results using MPEG-I video traces for predicting GOP rates show that the queue size is reduced by a factor of 15-160 and the network utilization is increased between 190%-300% as compared to a fixed service rate. Results also show that even when renegotiations occur on the average in tens of seconds, the queue size is reduced by a factor between 16-30     

Nonlinear prediction of a class of random processes

This paper is concerned with the minimum mean-squared error (MMSE) nonlinear prediction of a class of random processes. A class of random processes is defined by the property that its MMSE zero-memory predictor is represented by a finite sum of separable terms. Sufficient conditions for the existence of such processes are also considered. The nonlinear predictor is restricted to be composed of a linear filter in parallel with a zero-memory nonlinearity (ZNL) preceded by a variable delay, The optimum predictor is shown to be the solution of linear integral equations with the same kernel as for the optimum linear predictor. The first step of the derivation also yields a simpler scheme which ,only requires the addition of a ZNL to the optimum linear predictor. The improvements in the MMSE of the two nonlinear systems over the linear case are compared and illustrated by a numerical example.     

Nonlinear prediction in image coding with DPCM

In contrast to the traditional linear differential pulse code modulation (DPCM) design for the encoding of images, a new, nonlinear, neural network-based, DPCM technique has been devised. The predictor is designed by supervised training, based on a typical sequence of pixel values in an image. A function link neural network architecture has been used to design the predictor for one dimensional (1-D) DPCM. Computer simulation experiments in still image coding have shown that the resulting encoders work very well. At a transmission rate of 1 bit/pixel, for the image LENA, the 1-D neural network DPCM provides a 4.2 dB improvement in SNR over the standard linear DPCM system.<>     

Self-timed dynamically pipelined adaptive signal processing system: a case study of DLMS equalizer for read channel

Many pipelined adaptive signal processing systems are subject to a tradeoff between throughput and signal processing performance incurred by the pipelined adaptation feedback loops. In the conventional synchronous design regime, such throughput/performance tradeoff is typically fixed since the pipeline depth is usually determined in the design phase and remains unchanged in the run time. Nevertheless, in many real-life scenarios, the overall system performance can be potentially improved if we can run-time dynamically configure this tradeoff. With this motivation, we propose to apply self-timed pipeline, an alternative to synchronous pipeline, to implement the pipelined adaptive signal processing systems, in which the pipeline depth can be dynamically changed to realize run-time configurable throughput/performance tradeoffs. Based on a well-known high speed self-timed pipeline style, we developed architecture and circuit level design techniques to implement the self-timed pipelined adaptation feedback loop with configurable pipeline depth. We demonstrate the proposed design approach using a delayed least mean square (DLMS) adaptive equalizer for magnetic recording read channel. The data transfer rate in hard disk varies as the read head moves among tracks with different distance from the center of the disk platter. By adjusting the pipeline depth on-the-fly, the DLMS equalizer can dynamically track the best equalization performance allowed by the varying data transfer rates. Simulation result shows a significant performance improvement compared with its synchronous counterpart.     

自适应多码率语音编码流的可靠传输

自适应多码率语音编码已入选为第三代移动通信系统的语音压缩编码方案。本文提出了一种适合压缩语音传输的联合信源信道编码方法并对其性能进行了统计比较。利用压缩语音比特流中的固用冗余的信道译码算法是本文的研究重点。仿真结果表明利用信源冗余信息的信道译码器可以获得较大的编码增益。本文所用的信道编码方案为适合语音传输的卷积码。     

麦克风阵列的协同自适应滤波语音增强方法

在麦克风阵列语音增强方法中,传统的广义旁瓣抵消器在处理存在显著脉冲噪声的语音信号时效果较差。为提高在脉冲噪声干扰下的语音信号增强效果,提出一种麦克风阵列的协同自适应滤波语音增强方法。该方法采用协同自适应滤波取代线性自适应滤波,基于NLMS算法导出了滤波器权值和协同因子的自适应更新算法。仿真实验结果表明,所提方法能有效地消除掉语音信号的脉冲噪声和高斯噪声,克服线性自适应滤波对非线性脉冲噪声的不敏感性,比广义旁瓣抵消器效果优越很多。     

Nonlinear adaptive filtering of stimulus artifact

Noninvasive measurements of somatosensory evoked potentials have both clinical and research applications. The electrical artifact which results from the stimulus is an interference which can distort the evoked signal, and introduce errors in response onset timing estimation. Given that this interference is synchronous with the evoked signal, it cannot be reduced by the conventional technique of ensemble averaging. The technique of adaptive noise cancelling has potential in this regard however, and has been used effectively in other similar problems. An adaptive noise cancelling filter which uses a neural network as the adaptive element is investigated in this application. The filter is implemented and performance determined in the cancelling of artifact for in vivo measurements on the median nerve. A technique of segmented neural network training is proposed in which the network is trained on that segment of the record time window which does not contain the evoked signal. The neural network is found to generalize well from this training to include the segment of the window containing the evoked signal. Both quantitative and qualitative measures show that significant stimulus artifact reduction is achieved.     

A pseudo-differential MDAC with a gain-boosting inverter for pipelined ADCs

In this paper, a power-efficient pseudo-differential (PD) multiplying digital-to-analog converter (MDAC) is presented for pipelined analog-to-digital converters (ADCs). The proposed MDAC eliminates the explicit common-mode feedback circuit which is required in fully-differential configurations without any power penalty. Furthermore, a new class-AB gain-boosting inverter is proposed to be used in PD MDAC structures for further power saving. This inverter provides dynamic load current with no significant static power consumption and achieves high DC gain using a new gain-boosting technique. To demonstrate the effectiveness of the proposed circuits, they are utilized in the realization of a 1.5-bit/stage 10 bit 100 MS/s pipelined ADC.     

Nonlinear dynamic analysis of speech from pathological subjects

Nonlinear dynamic methods are employed to describe the complexity of speech from healthy and pathological subjects with vocal polyps. The analysis demonstrates the low-dimensional dynamic characteristics of normal and pathological voices as well as their statistically significant differences. The potential clinical application of nonlinear dynamics in speech signal processing of pathological voices is suggested     

Nonlinear noise cancellation for image with adaptive neuro-fuzzy inference systems

The adaptive neural-fuzzy inference system (ANFIS) algorithm is proposed for nonlinear noise cancellation of images. The quality in terms of mean square error of image restoration using the proposed ANFIS is over 65 times better for Gaussian noise and 500 times better for salt and pepper noise than using conventional filtering systems.