Harri Raittinen

Comparison of Center Estimation Algorithms for Heart and Respiration Monitoring With Microwave Doppler Radar

Microwave doppler radar offers significant improvements for unobtrusive heart and respiration measurement. Radar monitoring enables non-contact measurement, through clothing, of heart and respiration rate, which is desired in several applications ranging from medical sleep laboratory measurements to home health care measurements and stress monitoring. The use of high frequency radar (>; 10 GHz) instead of lower frequencies (~2.4 GHz) increases the signal-to-noise-ratio of the signal and enables the utilization of commercial radar modules. However, if high frequency radar is used, linear combining of quadrature radar channels is inadequate. Instead, a nonlinear channel combining algorithm is needed. The combining can be performed with an arctangent function if center, amplitude error, and phase error are estimated accurately and corrected. In this paper, we show that the Levenberg-Marquardt (LM) center estimation algorithm outperforms the state-of-the-art center estimation algorithm precision-wise and is computationally less complex. The simulated results show that the root mean squared error with the LM method is always less than 1%, while it is around 5%-13% with the compared method, depending on the breathing signal model used. In addition, the computational complexity of the LM method stays almost constant as the size of the data set increases, whereas with the reference method, it increases exponentially. In this paper, the LM method is validated both with simulations and with real data.

Image Deconvolution with Simulated Annealing Method

Simulated annealing method is a modern technique for solving various hard combinatorial optimization problems. It is based on principles drawn fom statistical physics and is quite ideal for NP-complete problems in which prior knowledge of the system behaviour is either scarce or lacking. This method has found applications in minimizing (or maximizing) functions with large numbers of parameters, e.g., in image restauration and VLSI floorplanning. Here we report a new image restauration method for getting rid of image blur, as caused by motion or out of focus. The reconstruction is performed by deconvolution, which we have found useful both for black and white and colour images.

VLSI Architecture of the Boltzmann Machine Algorithm

A new efficient programmable implementation of Boltzmann Machine algorithm will be presented. It is based on pulse-density modulation technique. Advantages of the design are simple structure of a synapse and thus small area, modularity and expandability. Furthermore, these structures can be used for various other neural network architectures. Applications for this type of networks can be found in the area of pattern recognition, image restauration and various optimization tasks.

Development of an Active Noise Cancellation Hearing Protector: How Can Passive Attenuation Be Retained?

The modest low-frequency attenuation of the conventional passive hearing protector can be improved electronically by active noise cancellation techniques. This paper presents the theory and some of the actual limitations of an active noise cancellation hearing protector. Three prototypes with similar types of controller, but with different mechanical construction were made and their performance was measured. The electronics of the system were implemented using analog electronics and feedback construction. The measurement results were compared with the results of an equivalent passive hearing protector with no internal electronics and with an intact earcup. The results show that the integration of the controller inside the earcups degrades the passive attenuation of the hearing protector at frequencies below 200 Hz. With proper design, an active noise cancellation hearing protector can still have 15 dB more noise attenuation at 100 - 200 Hz range than an equivalent passive hearing protector.

Fractal based image compression with affine transformations

As the needs for information transfer and storage increase, data coding and compression become increasingly important in applications such as digital HDTV, telefax, ISDN and image data bases. The authors have developed a fractal image compression method and tested it with binary (black and white) images. The decoded results are similar to the original images. The compression ratios are found to be extremely high.<<ETX>>

General purpose simulated annealing on hardware

A general purpose hardware for the Simulated Annealing optimization method has been designed and implemented. It is a digital stand-alone system that requires no host computer. The prototype design utilizes a full custom VLSI random number generator chip among standard digital circuits. Alteras reprogrammable FPGA technology is used to increase the packing density and to allow modifications. Thisfast optimization tool call be used to optimize any dynamically configurable hardware to adapt a given task. All application example and performance estimations are also presented.

ACTIVE NOISE CANCELLATION HEARING PROTECTOR WITH IMPROVED USABILITY

ABSTRACT Low frequency performance of a passive hearing protector can be improved significantly by active noise control methods. In a laboratory environment, the low frequency attenuation of a passive hearing protector can be improved by as much as 25 dB. However, when the device is used in real-life situations, several usability aspects must also be considered. A prototype of an active noise cancellation hearing protector was constructed. The main goal was to achieve comfortable and unnoticeable operation. Stability issues were also taken into account. The developed prototype is comfortable to use because of its improved stability and reduced nonlinear distortion.

Dynamically configurable combinatory logic array as Boolean neural network

A dynamically configurable combinatory logic array has been developed to be used as a Boolean neural network. The dynamical configuration facility enables hardware training using the simulated annealing method. The system is capable of adapting to a binary mapping task on the basis of given examples. The design target has been a cascadable VLSI chip.

SELF-ORGANIZING LOGIC USING SIMULATED ANNEALING

A selforganizing bit mapping neural network architecture using Simulated Annealing Method as optimization tool is presented. A graphic user interface for circuit simulator has been developed for simulation purposes. A flexible VLSI realization is also described.

Performance analysis of an active noise cancellation hearing protector

When designing an active noise cancellation hearing protector, it is relatively easy to implement a laboratory prototype with high performance. However, when the device is taken out of the laboratory into real extremely noisy situations, it must be capable of producing very high sound pressure levels. In this paper, the factors, which limit the low frequency performance and dynamic range are analysed. The results show that the voltage swing of the loudspeaker amplifier and the force factor of the loudspeaker are two important limiting factors in extremely loud situations. A prototype of an active noise cancellation hearing protector was implemented and its performance was measured in an authentic noisy situation. The developed device attenuated 125 Hz tonal noise by 18 dB. The active noise cancellation system attenuated noise even at a sound pressure level of 118 dB SPL. This can be considered sufficient for most situations.