A Pulsed Neural Network With On-Chip Learning and Its Practical Applications

Zhuang, Hualiang; Low, Kay Soon; Yau, Wei-Yun

doi:10.1109/tie.2006.888684

Cited by 31 publications

(22 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The most common activation function is tanh and the definition is shown below. (2) This activation function was also rejected. The Elliot function does approach one hyperbolically but not at the same rate as tanh and therefore is not interchangeable.…”

Section: Activation Functionmentioning

confidence: 99%

“…Such nonlinear transformations can be done efficiently using neural networks, but their practical implementation face another challenge. Until now neural networks are mostly implemented on computers with significant computational abilities to solve many types of real world problems [1][2][3][4]. Many people have put neural networks on FPGAs, DSPS, or high end embedded processors such as the ARM cores [2] [4].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Compensation of Sensors Nonlinearity with Neural Networks

Cotton

Wilamowski

2010

2010 24th IEEE International Conference on Advanced Information Networking and Applications

View full text Add to dashboard Cite

This paper describes a method of linearizing the nonlinear characteristics of many sensors using an embedded neural network. The proposed method allows for complex neural networks with very powerful architectures to be embedded on a very inexpensive 8-bit microcontroller. In order to accomplish this unique training software was developed as well as a cross compiler. The Neuron by Neuron process was as developed in assembly language to allow the fastest and shortest code on the embedded system. The embedded neural network also required an accurate approximation for hyperbolic tangent to be used as the neuron activation function.This process was then demonstrated on a robotic arm kinematics problem.

show abstract

Section: Activation Functionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Compensation of Sensors Nonlinearity with Neural Networks

Cotton

Wilamowski

2010

2010 24th IEEE International Conference on Advanced Information Networking and Applications

View full text Add to dashboard Cite

show abstract

“…The position dependent cogging force and friction force that have high nonlinearity are the main disturbance sources of PMLM, which can degrade the performance significantly. Many researchers have proposed various friction compensation methods [1][2][3][4][5][6][7][8]. The most frequently used method is the model based friction compensation approach [2][3][4].…”

Section: Introductionmentioning

confidence: 99%

“…Thus, it is difficult to obtain accurate parametric model. For this reason, several design methods without parametric model have been proposed [5][6][7][8].…”

Section: Introductionmentioning

confidence: 99%

Repetitive Model Predictive Control of a Precision Linear Motor Drive

Cao

Low

2007

IECON 2007 - 33rd Annual Conference of the IEEE Industrial Electronics Society

Self Cite

View full text Add to dashboard Cite

In this paper, a repetitive model predictive controller (RMPC) is proposed for tracking control of a linear motor. The proposed algorithm reduces the tracking error caused by the periodic disturbance by including the estimation of the future disturbances in the prediction model. The repetitive tracking error is treated as the consequence of periodic input disturbance, which is learnt over several trials and is used to improve the tracking accuracy. As compared with the conventional NPC and repetitive control, this algorithm attempts to derive the merits of the two schemes. Some experimental results are presented to demonstrate the effectiveness of the proposed approach.

show abstract

“…Thus t j can be utilised for a competitive learning algorithm for clustering. The output neuron v j is of the integrate-and-fire (IAF) type [7]. When its integrated potential s j (t) reaches the threshold u, an output pulse emits.…”

mentioning

confidence: 99%

On-chip pulse based parallel neural circuits for object-tracking system

2010

View full text Add to dashboard Cite

A scheme of pulse based neural circuits for object tracking is proposed. Different from the conventional frame-based methodology, the proposed design utilises parallel arrays of circuits to extract pixels with significant temporal contrast, which are indicators of moving objects. It can be implemented on an FPGA chip in full parallelism and improves the tracking performance dramatically. Moreover, its integrate-and-fire neural model can cluster concave data sets. Experimental results show that the proposed scheme outperforms conventional methods.Introduction: To automatically track multiple moving objects in a scene, conventional vision algorithms have to sequentially scan all pixels in each frame, and frame by frame. This leads to a tremendous computational burden if a high resolution camera is used, and it becomes a bottleneck for dynamic vision [1, 2]. Various methods have been proposed to cope with this issue. In [3], an optical flow algorithm is used to detect moving objects with a processing cycle time of 66 -166 ms. However, some frames may not be processed if a camera with a frame rate of higher than 25 fps is used. This would limit the response of the servo controller that is used to position the vision system. K-means algorithms are popular in clustering-based object tracking schemes [4]. The key weakness is that they cannot deal well with concave data sets in practical scenarios [5]. In [6], hardware improvement in the sensor layer is attempted to increase the processing speed. Lichtsteiner et al. propose a dynamic vision sensor (DVS), which is designed based on the addressed event representation (AER) technique and is different from the conventional frame based CCD and CMOS sensors. Although its dynamic performance is good, it has weakness in the target's detail recognition. In addition, its customised IC design is too specialised for most applications.

show abstract

A Pulsed Neural Network With On-Chip Learning and Its Practical Applications

Cited by 31 publications

References 8 publications

Compensation of Sensors Nonlinearity with Neural Networks

Compensation of Sensors Nonlinearity with Neural Networks

Repetitive Model Predictive Control of a Precision Linear Motor Drive

On-chip pulse based parallel neural circuits for object-tracking system

Contact Info

Product

Resources

About