Search

A method for the in-situ characterization of spurious modes in high power microwave transmission lines is presented. It uses hole couplers, which are integrated into a mitre-bend. The signals are proportional to the field strength at the mirror and allow to partly reconstruct the mode spectrum in the waveguide. The theoretical background is presented as well as results from measurements on prototypes.

Clustering of Gaussian mixture components, i.e. Hierarchical Gaussian mixture model clustering (HGMMC) is a key component of Gaussian selection (GS) algorithm, used in order to increase the speed of a Continuous Speech Recognition (CSR) system, without any significant degradation of its recognition accuracy. In this paper a novel Split-and-Merge (S&M) HGMMC algorithm is applied to GS, in order to achieve a better trade-off between speed and accuracy in a CSR task. The algorithm is further improved by introducing model selection in order to obtain the best possible trade-off between recognition accuracy and computational load in a GS task applied within an actual recognition system. At the end of the paper we discuss additional improvements towards finding the optimal setting for the Gaussian selection scheme.

In this paper, the impact of the pitch on the variability of MFCC, and their influence on the performance of the automatic speech recognition system, is analyzed. In case that a speaker has a high pitch, the distance between adjacent harmonics in the spectrum of voiced phonemes is larger, which results in poorer description of the spectral envelope. Additional problem arises in the case that a band-pass filter from the analysis filter bank covers the range between two harmonics, capturing the part of spectrum without energy, which consequently leads to the detection of sudden, non-existing changes in the spectral envelope. The reduction of these variations is analyzed by using a lower number of MFCC, by expending the bandwidths of the band-pass filters from the filter bank at lower frequencies, and also by low-pass filtering of the filter bank output. Some of the results are somewhat different from the similar results presented in the literature, for which the adequate explanations are offered.

We developed a method for automatic detection of the optimal contact within a multi-pad surface electrode that could be used for wrist tremor reduction in individuals with Parkinson's Disease or Essential Tremor. This method is based on custom designed program that determines the optimal contact position, i.e. the position on the skin that is closest to the flexor/extensor muscle motor point. The method also takes into account patients' comfort, as the contacts that produce painful contractions or stimulate sensory pathways are eliminated from the selection process.

Neural prostheses (NP) for tremor reduction rely on the information of tremor frequencies and their intensity and timing. Better estimation of tremor is important for reliable input to NP. In this work we compare two methods for tremor extraction. Discrete wavelet transform and Butterworth filter were used to decompose pathological volitional forearm movement into its slow and fast components. Slow signal is associated with volitional motion while the fast rhythmic signal is the result of involuntary muscular contraction (tremor). Our results suggest that in absence of prior knowledge about tremor frequency characteristics, wavelet filtering has advantage over other conventional filtering methods.

The paper presents the first monolithic, microstrip-based X-band tuner with ohmic contact MEMS switches for efficiency optimization of class-E power amplifiers (PAs). The tuner is based on a "3 bit-3 bit" digital design, and uses 6 radial shunt stubs that can be selected via the integrated ohmic contact MEMS switches. The simulation tuning range for post-production efficiency optimization of a 10 GHz amplifier varies from 0.63 /spl Omega/ to 57 /spl Omega/ for the real part and from -7.5 /spl Omega/ to 51.8 /spl Omega/ for the imaginary part. Measured results of the monolithic tuners fabricated on a silicon substrate show good agreement with the simulation tuning range of the PAs. The measured maximum VSWR is 38, and the tuner size is 6/spl times/7 mm/sup 2/.

We present a monolithic 60-GHz quasi-optical amplifier antenna array. An input plane wave is received by the slot antennas in the ground plane, capacitively coupled through the GaAs substrate to the microstrip PHEMT amplifiers, and radiated in transmission by patch antennas.

We present several broad band arrays of tapered slot antennas for use in a new generation phased array telescope. In specific, a four element linearly polarized 3-12 GHz array, a 3 element 0.8-2.5 GHz coupled array and a four element dual polarized 0.8-2.5 GHz array are discussed. Integration with LNAs and coupling between elements is examined.