December 1999, Volume 8, Number 4
This paper presents new Czech language two-channel (stereo) speech database recorded in car environment. The created database was designed for experiments with speech enhancement for communication purposes and for the study and the design of a robust speech recognition systems. Tools for automated phoneme labelling based on Baum-Welch re-estimation were realised. The noise analysis of the car background environment was done.
The state-space description of digital filters involves except the relationship between input and output signals an additional set of state variables. The state-space structures of digital filters have many positive properties compared with direct canonical structures. The main advantage of digital filter structures developed using state-space technique is a smaller sensitivity to quantization effects by fixed-point implementation. In our presentation, the emphasis is on the analysis of coefficient quantization and on existence of zero-input limit cycles in state-space digital filters. The comparison with direct form II structure is presented.
In non stationary channels, error rates vary considerably. This paper proposes an effective go-back-N (GBN) Automatic-Repeat-Request (ARQ) scheme which estimates the channel state in a simple manner, and adaptively switches its operation mode. It provides higher throughput than other comparable ARQ schemes in conditions of Land-Mobile-Satellite (LMS) channel.
This submission dealt with the design, simulation and realisation of a dual mode oscillator with distributed RC structures that use AT-cut quartz resonator as stabilisation element and sensor element of its own temperature. Simulations, experimental results and evaluation of the short-term stability are also included.
This paper deals with methods of speech enhancement with particular focus on neural speech enhancement. Speech enhancement is concerned with the neural processing of noisy speech to improve the quality and intelligibility of the speech signal. The goal of this paper is to describe an experiment with implementation of two channel adaptive noise canceler via direct time domain mapping approach.
The paper presents basic equations of efficient GSM Viterbi equalizer algorithm based on approximation of GMSK modulation by linear superposition of amplitude modulated pulses. This approximation allows to use Ungerboeck form of channel equalizer with significantly reduced arithmetic complexity. Proposed algorithm can be effectively implemented on the Viterbi and Filter coprocessors of new Motorola DSP56305 digital signal processor. Short overview of coprocessor features related to the proposed algorithm is included.
A novel method for analogue high-level fault simulation (HLFS) using linear and non-linear high-level fault models is presented. Our approach uses automated fault model synthesis and automated model selection for fault simulation. A speed up compared with transistor-level fault simulation can be achieved, whilst retaining both behavioural and fault coverage accuracy. The suggested method was verified in detail using short faults in a 10k state variable bandpass filter.
This paper presents the performance of four motion estimation algorithms with present of zero-mean Gaussian noise. The performance of full search (FS) algorithm, three-step search (TSS) algorithm, new three-step search (NTSS) algorithm, and four-step search (4SS) algorithm with influence of Gaussian noise has been described for three of famous video sequences. The critical signal-to-noise ratio (SNR) is introduced for each algorithm. The worst performance is often in case of full search algorithm (FS).