H. Kwan

Learn More
UltraSPARC is the first microprocessor from Sun Microsystems' SPARC Technology Business to implement the new 64-bit SPARC V9 architecture. ULtraSPARC is equipped with unique multimedia capabilities and is capable of 4-way superscalar instruction dispatch, with an emphasis on maximal system efficiency and throughput in the execution of complex,(More)
This paper presents an adaptive neuro-fuzzy filtering scheme using the artificial neuro-fuzzy inference system (ANFIS) for noise reduction in speech. The measurable output noisy speech with 5 dB SNR level is taken as the contaminated version of the interference to compare with the output data of the filter. The white noise source is taken as the input. With(More)
A Levinson-type algorithm is proposed for estimating the two-dimensional autoregressive-moving average (ARMA) model from the observed inputs and outputs of an unknown 2-D system. By applying the concept of dummy prediction errors, the algorithm can be extended for computing 2-D models of any AR-order and MA-order. It is shown that the algorithm can be(More)
A real Gabor transform (GT) for image analysis and compression is presented in this paper and the RLS algorithm is applied to extract the coefficients of the real 2-D GT. A significant computation of the real GT can be saved when compared to the computation of the widely used complex Gabor transform. In addition, the real GT has a simple relationship with(More)
In this paper, the use of the ARMA lattice model for glottal source excited pitch synchronous speech analysis and synthesis is presented. By modeling the vocal tract with the ARMA lattice model, the effect of the nasal cavity is modeled by the introduced zeros. Consequently, a higher quality speech could be synthesized. In so doing, the limitation due to(More)
In this paper, a 2-D phoneme sequence pattern recognition using the fuzzy neural network is presented. The self-organizing map and the learning vector quantization are used to organize the phoneme feature vectors of short and long phonemes segmented from speech samples to obtain the phoneme maps. The 2-D phoneme response sequences of the speech samples are(More)
A bit-level systolic array is presented for the implementation of a digital sinusoidal oscillator with low-frequency and low-amplitude sensitivities and low-roundoff noise. The oscillator is derived from a lossless digital two-port and is extremely suitable for low-frequency applications where conventional direct form and coupled form oscillators fail to(More)
  • H.K. Kwan
  • 2005
In this paper, the construction and design of analysis and synthesis filterbanks using analysis sub-filters that consist of second-order passive digital filters, and synthesis sub-filters that consist of low-order finite impulse response (FIR) digital filters are presented. The use of low-order FIR digital filters reduces the group delay of the analysis and(More)
A novel neural network method for adaptive IIR digital filtering is proposed. Based on the linear prediction principle and the observable data at the IIR filter input and output prior to the current iteration time k, an analog neural network is used to estimate the filter coefficients of the next iteration time k+1. Computer simulation results are given(More)
Two systolic structures for the realization of two-dimensional FIR linear-phase decimators and interpolators are presented. The structures are derived in terms of decimators only since an interpolator structure can be obtained by a direct transposition of the signal flow graph of the corresponding decimator. One of the systolic structures is found to have(More)