Bank Of Bandpass Filters Research Articles

Noise-Robust Pitch Detection using Auto-correlation Function with Enhancements

An efficient noise-robust pitch detection algorithm is proposed in this paper. The algorithm is based on time domain autocorrelation function (ACF). A bank of band-pass filters is used for competitive contribution of periodicity to select primary pitch candidates. A weighting criterion that involves both increase and decrease in merit is applied to the candidates by exploiting the presence or the absence of pitch harmonics. Finally, a simple enhancement is integrated to smooth the pitch contour. The proposed algorithm is evaluated on TIMIT database in different types and levels of noise in terms of pitch and voice activity detection. The experimental results show the superiority of the proposed method over well known other methods.

Development of multichannel intermediate frequency system for electron cyclotron emission radiometer on KSTAR Tokamak

Plasma experiments on KSTAR are scheduled to start up this year (2008). We have developed an electron cyclotron emission (ECE) radiometer to measure the radial electron temperature profiles in KSTAR experiments. The radiometer system consists, briefly, of two downconversion stages, amplifiers, bandpass filter banks, and video detectors. These components are made commercially or developed in house. The system detects ECE power in the frequency range from 110 to 196 GHz, the detected signal being resolved by means of 48 frequency windows. Before installation of this system on KSTAR, we installed a part of this system on large helical device (LHD) to study the system under similar plasma conditions. In this experiment, the signal amplitude, considered to be proportional to the electron temperature, is measured. The time-dependent traces of the electron temperature measured by this radiometer are in good agreement with those provided by the LHD Michelson spectrometer. The system noise level which limits the minimum measurable temperature (converted to the electron temperature) is about 30 eV.

Development of microwave imaging reflectometry in large helical device

Three key devices of the microwave imaging reflectometry (MIR) are under development in large helical device (LHD). The 2-D mixer array is developed by stacking the one-dimensional array of the planar Yagi-Uda antenna. The new type of the bandpass filter bank is modified to match the requirement of the MIR. The low-cost quadrature demodulator is also developed for the phase detection system. By using the low-price commercial wireless devices, the development cost becomes much lower than the expensive waveguide system. These devices enable the development of 2-D/3-D microwave imaging system for the plasma diagnostics and industrial applications.

Inharmonic dispersion tunable comb filter design using modified IIR band pass transfer function

An excitation/filter system of Inharmonic sound synthesis signal is presented with an application to piano, a stiff string instrument. Specific features of the piano string important in wave propagation, dispersion due to stiffness, frequency dependent losses and presence of phantom partials are included in the proposed model. The modified narrow bandpass filter is used as a basic building block in modeling the vibrating structure. The parallel bank of narrow band pass filters is used to model the dispersion. The center frequencies of narrow bandpass filters can be tuned to follow the same relation as the partial frequencies of piano tone. Novel loss filter is designed to implement frequency dependent losses. The resulting model is called as Inharmonic Dispersion Tunable Comb filter.

Considering a Non-Uniform Filter Bank in an UWB Multi Band On-Off Keying Transceiver

Ultra Wide Band (UWB) wireless systems based on impulse radio have the potential to provide very high data rates over short distances. One of the solutions for the UWB communication systems is the Multi Band On-Off Keying (MB-OOK) which consists in an OOK modulation generalized over multiple frequency sub-bands and associated with a demodulation based on a non-trivial energy threshold comparison. This article presents the analysis and simulations of an UWB MB-OOK transceiver in the 6-8.5GHz band. This study is focused on the bandpass filter bank in the transceiver’s architecture. The use of a Bulk Acoustic Wave (BAW) filter bank in the transceiver is considered. Finally, system performance is estimated for the proposed architecture.

Real time decomposition of speech into modulated components

Motivated by the active process of the outer hair cell (OHC) in the mammalian inner ear, a real time decomposition of speech into modulated components is presented. A generalized phase lock loop (GPLL) was applied to decompose the speech signal into its envelope and positive instantaneous frequency (PIF) parts, which can be further processed and represented by timing information alone. A log-derivative operator is applied to the bandpass signal. Analytic and antianalytic components occupying non overlapping frequency bands are separated by filtering. The proposed algorithms are used to represent speech signals processed through a bandpass filter bank.

Hand Motion Gesture Frequency Properties and Multimodal Discourse Analysis

Gesture and speech are co-expressive and complementary channels of a single human language system. While speech carries the major load of symbolic presentation, gesture provides the imagistic content. We investigate the role of oscillatory/cyclical hand motions in `carrying' this image content. We present our work on the extraction of hand motion oscillation frequencies of gestures that accompany speech. The key challenges are that such motions are characterized by non-stationary oscillations, and multiple frequencies may be simultaneously extant. Also, the duration of the oscillations may be extended over very few cycles. We apply the windowed Fourier transform and wavelet transform to detect and extract gesticulatory oscillations. We tested these against synthetic signals (stationary and non-stationary) and real data sequences of gesticulatory hand movements in natural discourse. Our results show that both filters functioned well for the synthetic signals. For the real data, the wavelet bandpass filter bank is better for detecting and extracting hand gesture oscillations. We relate the hand motion oscillatory gestures detected by wavelet analysis to speech in natural conversation and apply to multimodal language analysis. We demonstrate the ability of our algorithm to extract gesticulatory oscillations and show how oscillatory gestures reveal portions of the multimodal discourse structure.

Human Hearing Modelling Real-Time Spectrography for Visual Feedback in Singing Training

The traditional form of spectrography employed for speech analysis makes use of fixed bandwidth bandpass filters, which are usually set either to wide-band or narrow-band to optimise the time and frequency resolution of the resulting spectrogram respectively. The acoustic analysis carried out by the human ear can be modelled as a bank of bandpass filters whose bandwidth changes as a function of centre frequency, and therefore the time and frequency resolution varies with frequency. This paper describes a spectrographic analysis technique that is based on peripheral human hearing and considers its potential for application within a real-time visual display for singing training by comparing its output with traditional wide- and narrow-band spectrograms. The potential advantages and disadvantages of hearing modelling spectrography for this application are illustrated and discussed for a selection of sung material, some of which has been recorded during singing lessons where traditional spectrography is being employed.

Evaluation of an auditory model for echo delay accuracy in wideband biosonar.

In a psychophysical task with echoes that jitter in delay, big brown bats can detect changes as small as 10-20 ns at an echo signal-to-noise ratio of approximately 49 dB and 40 ns at approximately 36 dB. This performance is possible to achieve with ideal coherent processing of the wideband echoes, but it is widely assumed that the bat's peripheral auditory system is incapable of encoding signal waveforms to represent delay with the requisite precision or phase at ultrasonic frequencies. This assumption was examined by modeling inner-ear transduction with a bank of parallel bandpass filters followed by low-pass smoothing. Several versions of the filterbank model were tested to learn how the smoothing filters, which are the most critical parameter for controlling the coherence of the representation, affect replication of the bat's performance. When tested at a signal-to-noise ratio of 36 dB, the model achieved a delay acuity of 83 ns using a second-order smoothing filter with a cutoff frequency of 8 kHz. The same model achieved a delay acuity of 17 ns when tested with a signal-to-noise ratio of 50 dB. Jitter detection thresholds were an order of magnitude worse than the bat for fifth-order smoothing or for lower cutoff frequencies. Most surprising is that effectively coherent reception is possible with filter cutoff frequencies well below any of the ultrasonic frequencies contained in the bat's sonar sounds. The results suggest that only a modest rise in the frequency response of smoothing in the bat's inner ear can confer full phase sensitivity on subsequent processing and account for the bat's fine acuity or delay.

Measurement of short reverberation times at low frequencies using wavelet filter bank

In room acoustics, reverberation time is an important acoustic parameter. However, it is often difficult to measure short reverberation times at low frequencies with a traditional band pass filter bank if the product of filter bandwidth (B) and reverberation time (T) is small. It is well known that the minimum permissible product of bandwidth and reverberation time of the traditional band pass filter is at least 16. This strict requirement makes it difficult to measure short reverberation times of an acoustic room at low frequencies exactly. In order to reduce this strict requirement, in the previous paper, the wavelet filter bank was developed and the minimum permissible product of bandwidth and reverberation time was replaced with 4. In the present paper, it is demonstrated how the short reverberation times of an practical room at low frequencies are successfully measured by using the wavelet filter bank and the results are compared with the traditional method using a band pass filter bank.

Speaker recognition using dynamic synapse based neural networks with wavelet processing

We have designed systems utilizing a dynamic synapse neural network (DSNN) to perform speech recognition tasks. The DSNN architecture has demonstrated noise-resistive properties in previous research. The speech signal is first passed through a bank of bandpass filters. The output of each filter then undergoes wavelet transformation to extract speaker-dependent features. Transformed signals are used as input for a DSNN, which is trained using a genetic algorithm training method. We are seeking to explore the capabilities of a biologically based neural network in solving a problem which humans are still better at solving than conventional solutions. The dynamic synapse neural networks were trained to perform several speech recognition tasks, including speaker verification, speaker identification, and word recognition. In the speaker verification task, the objective was to distinguish one speaker from a group of up to 22 imposters based on the utterance of a codeword. In the speaker identification task, the objective was to identify which of 6 speakers has uttered the codeword. In the word recognition task, the objective was to perform speaker-dependent word recognition on a set of 30 words. Neural networks were capable of 90% correct identifications in the presence of white noise (0 dB SNR).

Genome wide oscillations in expression. Wavelet analysis of time series data from yeast expression arrays uncovers the dynamic architecture of phenotype.

A reanalysis of expression arrays in yeast cells synchronized by alpha factor blockade or through the use of temperature sensitive mutants uncovered a genome wide pattern of oscillations in mRNA concentrations. Using wavelet decomposition as a signal processing technique and enhancement strategies borrowed from image processing, noise and trends in the Stanford yeast cell cycle data were partitioned away from time series profiles to uncover genome-wide oscillations in expression. These oscillations which were typically of cell cycle or half cell cycle duration, 40 and 80 minutes in the Stanford data set suggest that there are large-scale temporal structures and high frequency oscillations in mRNA levels through the cell cycle. Wavelet decomposition, which acts like a band pass filter bank, was used to determine where most of the power appeared in the decomposition. The approximately 40-min oscillation is mirrored in continuous chemostat cultures. In these cultures, metabolic synchrony involving an unknown proportion of the transcriptome can be monitored by measurement of oxygen consumption and can be sustained for weeks. These 40-min oscillations are stable and precise with coefficients of variation less than 1% for both period and amplitude. The hypothesis that high and low amplitude oscillations are a ubiquitous property of the genetic regulatory circuitry was supported by the observation of period doubling bifurcations in the distribution of population doubling times in yeast.

OPTIMIZING MODELS OF DOLPHIN AUDITORY SENSITIVITY USING EVOLUTIONARY COMPUTATION

ABSTRACT Object classifiers that attempt to mimic dolphin echolocation require an auditory weighting function representative of dolphin peripheral auditory processing. An evolutionary program (EvPg) was used to fit the frequency-dependent output of a bank of bandpass filters to the auditory sensitivity of the bottlenose dolphin, Tursiops truncatus. Pseudo-Gaussian (PG) and rounded exponential (ROEX) functions were used to describe individual filter shapes. Variables determining the number of filters per model, overall filter shape and amplitude scaling were submitted to the EvPg for optimization. Maximum deviation (P e ) between model output and the sensitivity of the dolphin was used as a measure of similarity between the two, i.e., lower P e indicated a greater similarity. The number of filters converged upon 37 for all ROEX models and ≤ 45 for all PG models. The P e of the best-performing PG model was 0.08, and for all ROEX models was 0.13. Greatest deviations typically occurred below 5 kHz and above 130 kHz. Relative audiometric sensitivity of a dolphin ear model has been improved relative to previous models, thereby providing an auditory weighting function more representative of dolphin peripheral auditory processing. This model will be applied to further investigate how dolphins use echolocation to discriminate among objects.

On detection using filter banks and higher order statistics

We suggest a method, based on the use of filter bank and higher order statistics (cumulants), for detection of transient signals. The method first uses a bandpass filter bank, which separates the spectrum of the observed signal into narrow frequency bands. Each subfilter of the filter bank is then followed by a cumulant estimator, and thereby suppressing colored noise. By selecting those subfilters that have large output energies, the filter bank can approximate the behavior of a matched filter. Moreover, no a priori information about the waveform of the signal is needed. The performance of the detector is evaluated by using a simulated signal as well as a measured signal. The presented detector is compared with the optimal matched filter detector.

Frequency-division-multiplexed polarimetric fiber laser current-sensor array

We demonstrate a novel frequency-domain-multiplexing technique for implementing polarimetric fiber laser current sensors. Each sensor operates at a different polarization-mode beat frequency that is modulated in response to applied electric current. A bank of bandpass filters can be used to separate signals from different sensors. A simple frequency-demodulation technique based on a phase-locked loop is used for signal processing.

Modeling the effects of peripheral compression on temporal processing

Most psychoacoustic models to date have assumed that the peripheral auditory system can be treated as a bank of quasilinear bandpass filters. Physiological studies have shown that this approximation holds only for damaged cochleae: the response of the basilar membrane in a healthy cochlea to sounds around the characteristic frequency is highly nonlinear and compressive over a wide range of sound-pressure levels. Many aspects of psychoacoustic performance can be explained by taking these nonlinear characteristics into account. Short-term temporal integration in simultaneous and forward masking (where signal detectability improves with increasing signal duration), the critical masking interval (where the detectability of a brief signal temporally centered in a masker decreases with increasing masker duration), and the nonlinear growth of forward masking can be modeled with relatively simple assumptions if a physiologically reasonable nonlinearity is included. Similarly, some aspects of temporal processing in listeners with cochlear hearing impairment can be well described by making the reasonable assumption that cochlear compression is reduced or absent. Thus the altered temporal processing shown in some tasks by hearing-impaired listeners may in many cases be due to reduced cochlear compression and not to more central processing deficits. [Work supported by the NIH/NIDCD.]

Using Spectral Subtraction for Suppression of Noise in Speech Signals with Analog Integrated Circuits

A new modification of the spectral subtraction algorithm is presented which enables operating entirely in the time domain and is thus suitable for realization in analog integrated circuits. The noise spectrum is obtained during speechless intervals and stored for spectral subtraction when speech is present in the signal. The frequency range of interest of the speech signal is divided into narrow frequency bands by means of a bank of band-pass filters. For each frequency band the noise model is realized as an auxiliary signal multiplied by a particular weight. A subsystem is presented that produces an output signal whose power is equal to the difference between the input signal power and the noise model power for each frequency channel, thereby realizing the spectral subtraction. Circuits to achieve the described operation are outlined. Finally, simulation results of the noise removal algorithm are shown in the form of a spectrogram and the results showing improvement in automatic speech recognition are given.

Interaural-delay sensitivity to frequency modulation

Data are presented on experiments that compare binaural sensitivity to sinusoidal amplitude and frequency modulation (SAM and SFM). Among the stimulus parameters examined are carrier frequency (3–8 kHz), modulation rate (25–800 Hz), and depth (β=0.1–2). Observers are most sensitive to interaural delays in SAM tones at rates of about 50 to 300 Hz, while for SFM tones, best sensitivity occurs at rates of 200 to 400 Hz. Interaural-delay thresholds for SFM tones are quantitatively predicted from a stochastic AM-envelope model whose parameters are uniquely determined from data on SAM tones. The model uses short-term cross-correlation analysis of Hilbert envelopes at the outputs of a bank of peripheral bandpass filters. The envelopes are subsequently convolved with the impulse response of a low-pass FIR filter, followed by additive noise, and a decision device whose output is a position estimate. The expected value and variance of estimates are stochastically approximated, from which the model adaptively tracks the 0.707 probability of a correct response (d′=0.78) on the observer’s forced-choice psychometric function. [Work supported by NIH.]

Modeling the perception of concurrent vowels: Role of formant transitions.

When two synthetic vowels are presented concurrently and monaurally, listeners identify the members of the pair more accurately if they differ in fundamental frequency (F0), or if one of them is preceded or followed by formant transitions that specify a glide or liquid consonant. However, formant transitions do not help listeners identify the vowel to which they are linked; instead, they make the competing vowel easier to identify. One explanation is that the formant transition region provides a brief time interval during which the competing vowel is perceptually more prominent. This interpretation is supported by the predictions of two computational models of the identification of concurrent vowels that (i) perform a frequency analysis using a bank of bandpass filters, (ii) analyze the waveform in each channel using a brief, sliding temporal window, and (iii) determine which region of the signal provides the strongest evidence of each vowel. Model A [Culling and Darwin, J. Acoust. Soc. Am. 95, 1559-1569 (1994)] computes the rms energy in each channel at successive time intervals to generate running excitation patterns that serve as input to a vowel classifier, implemented as a linear associative neural network. Model B uses a temporal analysis in each channel to generate running autocorrelation functions, and it includes a further stage of source segregation [Meddis and Hewitt, J. Acoust. Soc. Am. 91, 233-245 (1992)] to partition the channels into two groups, one group providing evidence of the periodicity of the vowel with the dominant F0, the other group providing evidence of the competing vowel. Both models predicted effects of F0 and formant transitions on identification, but model B provided more accurate predictions of the pattern of listeners' identification responses. Taken together, the empirical and modeling results support the idea that the identification of concurrent vowels involves an analysis of the composite waveform using a sliding temporal window, combined with a form of F0-guided source segregation.

Speech formant frequency and bandwidth tracking using multiband energy demodulation

In this paper, the amplitude and frequency (AM–FM) modulation model and a multiband demodulation analysis scheme are applied to formant frequency and bandwidth tracking of speech signals. Filtering by a bank of Gabor bandpass filters is performed to isolate each speech resonance in the signal. Next, the amplitude envelope (AM) and instantaneous frequency (FM) are estimated for each band using the energy separation algorithm (ESA). Short-time formant frequency and bandwidth estimates are obtained from the instantaneous amplitude and frequency signals; two frequency estimates are proposed and their relative merits are discussed. The short-time estimates are used to compute the formant locations and bandwidths. Performance and computational issues of the algorithm are discussed. Overall, multiband demodulation analysis (MDA) is shown to be a useful tool for extracting information from the speech resonances in the time–frequency plane.