Abstract
Although voiced speech signals are physical signals which are approximately harmonic and electric power signals are true harmonic, the algorithms used for harmonic analysis in electric power systems can be successfully used in speech processing, including in speech enhancement, noise reduction, speaker recognition, and hearing aids. The discrete Fourier transform (DFT), which has been widely used as a phasor estimator due to its simplicity, has led to the development of new DFT-based algorithms because of its poor performance under dynamic conditions. The multiple-resonator (MR) filter structure proposed in previous papers has proven to be a suitable approach to dynamic harmonic analysis. In this article, optimized postprocessing compensation filters are applied to obtain frequency responses of the transfer functions convenient for fast measurements in dynamic conditions. An optimization design method based on the constrained linear least-squares (CLLS) is applied. This way, both the flatness in the passband and the equiripple attenuation in the stopband are satisfied simultaneously, and the latency is reduced.
Highlights
Signals that consist of a sum of sine waves whose frequencies are integral multiples of the lowest frequency are said to be harmonic
Many different approaches are possible; some basic underlying principles of speech enhancement techniques capitalize on the observation that waveforms of voiced sounds are periodic, with a period that corresponds to the fundamental frequency
Two following tests are given to illustrate the filter’s dynamic features, while static characteristics are obviously clear from the frequency responses
Summary
Signals that consist of a sum of sine waves whose frequencies are integral multiples of the lowest frequency (so-called fundamental) are said to be harmonic. Examples include voiced speech and other biological signals, musical waveforms, helicopter and boat sound waves, and outputs of nonlinear systems excited by a sinusoidal input [1,2]. Many different approaches are possible; some basic underlying principles of speech enhancement techniques capitalize on the observation that waveforms of voiced sounds are periodic, with a period that corresponds to the fundamental frequency. There are many signal analyses and transformations that are best performed in the frequency domain [6,7,8,9,10,11,12,13,14]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have