András Radványi

A digital multiprocessor hardware accelerator board for cellular neural networks : CNN-HAC

Analogue realizations of neural networks are superior in speed. the hardware accelerator boards using catalogue programmable VLSI ICs represent a trade-off having higher reconfigurability and lower cost. This paper presents such a solution for a cellular neural network (CNN). The architecture of the present design (CNN-HAC) using four standard DSPs to calculate the transient response of a one-layer CNN containing (0.25–0.75) × 106 analogue neural cells (depending on the type of template) is presented. the architecture and also the design principles are independent of the number of processors. the actual design was made in the form of a PC add-on board. The global control unit, which connects the board to the host firmware and communicates control signals to/from the local control units of the DSPs, was realized mainly with EPLDs. A special correspondence between the virtual processing elements—calculating the time-discrete models of the analogue neural cells—and the physical ones is discussed in detail. It is realized in an architecture with a simple, two-directional interprocessor communication. This architecture can be ‘scaled down’ using faster processors, EPLDs and memories. the present version runs with 2 μs/cell/iteration speed.

Detecting moving and standing objects using cellular neural networks

The general framework of motion detection based on discrete time samples of the moving image is defined. Four types of motion detection problem are studied. the simplest one is a model resembling the famous Hubel-Wiesel experiment with a cats retina for detecting the motion of an object having a given speed in a given direction. the most complicated case is the determination of the vertical and horizontal velocity components of a moving image. Various cloning template sequences are proposed for detecting different types of motion. In the sampled mode the consecutive black-and-white snapshots are fed to the input and to the initial state nodes of the cellular neural network respectively. After the transients have decayed, the output gives the information necessary for detecting the presence or absence of a specific motion as well as for estimating the direction and magnitude of the velocity vector. In continuous mode the sampling process is eliminated by the use of delay-type templates. Conditions are analysed under which the detection is correct. the circuit realization of some motion detectors is discussed and the use of a programmable dual-CNN structure is proposed.

On a CNN chip-prototyping system

An analogic CNN chip prototyping and development system was designed and manufactured to test and measure different VLSI implementations of the analogic CNN Universal Machine. A high level language was developed to support the design of analogic algorithms and an image capture was designed for on-chip image sensing and through CCD camera.<<ETX>>

A hardware accelerator board for cellular neural networks: CNN-HAC

The hardware accelerator (HAC) boards using catalog programmable VLSI ICs represent a trade-off having higher reconfigurability and lower cost. This paper presents such a solution for a cellular neural network (CNN). The architecture of the present design (CNN-HAC) using 4 standard DSPs to calculate the transient response of a one-layer CNN containing 0.25-1.0 million analog neural cells is presented. The architecture and also the design principles are independent of the number of processors. The actual design was made in the form of a PC add-on board. The global control unit, which connects the board to the host firmware and communicates control signals to/from the local control units of the DSPs, was realized mainly with EPLDs. A special correspondence between the virtual processing elements-calculating the time discrete models of the analog neural cells-and the physical ones, established to work an architecture with an infrequent, one-directional interprocessor communication, is discussed in detail.<<ETX>>

Analogic CNN algorithms for 3D interpolation‐approximation and object rotation using controlled switched templates

Analogic CNN algorithms are presented for various interpolation and approximation tasks in 3D. They are designed on the basis of mechanical analogies. Symmetric space-variant operations are implemented by the CNN algorithms; with switched templates, a key example is object rotation. Direction and speed coding are shown in detail.

On the rectangular grid representation of general CNN networks

Although the cellular neural paradigm in its original form provides a suitable framework for investigating problems defined on arbitrary regular grids, the chips—ready ones or under design—as well as the available simulators are all restricted to a rectangular structure. It is not at all self-evident, however, that the rectangular structure is the most suitable to represent every practical problem. In this paper we demonstrate that several cellular neural networks of various regular grids can be mapped onto the typical eight-neighbour rectangular one, by applying weight matrices of periodic space variance. By adopting this option, the applicability of cellular neural chips and simulators can be extended to investigate and solve problems of essentially arbitrary grid structures. Copyright

ACE: a digital floating point CNN emulator engine

The architecture of ACE, a multiprocessor analogic cellular neural network (CNN) emulator engine consisting of 2 to 16 TMS320C40 floating point DSPs is introduced. The engine containing up to 512 Mbyte RAM (enough to store a 512/spl times/512/spl times/512 sized CNN cube) which can be controlled through its SCSI port. It can either accelerate the multilayer CNN simulator CNNM or be accessed directly from the high level, C-based analogic CNN language ACL to achieve the simulation speed of /spl sim/2.8 /spl mu/sec/cell/iteration/DSP for 3/spl times/3 linear templates.

SPATIAL DEPTH EXTRACTION USING RANDOM STEREOGRAMS IN ANALOGIC CNN FRAMEWORK

Demonstrated and motivated on human stereo vision analogic CNN algorithms are proposed to extract 3D spatial information from computer-generated random-dot stereograms as well as real scene random-dot like ones produced with simple optical devices, projector and camera. Several aspects of making real scene stereograms are considered to minimize perspective distortion and enable local CNN processing.

Computational auditory scene analysis in cellular wave computing framework

Extracting substantive auditory objects from an auditory mixture is a widely studied and difficult problem in auditory research. Computational auditory scene analysis (CASA) aims at extracting objects in the frequency domain, based on a set of psychoacoustic grouping procedures. To mimic some aspects of the human auditory system a new cellular neural/non-linear network (CNN)-based library of procedures forming an Auditory Wave Computing Toolkit (AWCT) is presented here. Copyright

Using Cellular Neural Network to See Random-Dot Stereograms

The Cellular Neural Networks (CNN) providing for efficient analog array processing of images are used for revealing surface features hidden in different types of random-dot stereograms (RDS) coding 3D information in internal correlation. Although originally random-dot stereograms are for probing human stereopsis, they may also find application in engineering practice based on the method proposed for creating stereograms of real objects in optical environment using projector and camera. Also the concept of difference surface and difference stereogram is introduced and used for coding smooth surfaces.