Output Of Filter Bank Research Articles

Effects of Vibratory Source on Auditory-Perceptual and Bio-Inspired Computational Measures of Pediatric Voice Quality

The vibratory source for voicing in children with dysphonia is classified into three categories including a glottal vibratory source (GVS) observed in those with vocal lesions or hyperfunction; supraglottal vibratory sources (SGVS) observed secondary to laryngeal airway injuries, malformations, or reconstruction surgeries; and a combination of both glottal and supraglottal vibratory sources called mixed vibratory source (MVS). This study evaluated the effects of vibratory source on three primary dimensions of voice quality (breathiness, roughness, and strain) in children with GVS, SGVS, and MVS using single-variable matching tasks and computational measures obtained from bio-inspired auditory models. A total of 44 dysphonic voice samples from children aged 4-11years were selected. Seven listeners rated breathiness, roughness, and strain of 1000-ms /ɑ/ samples using single-variable matching tasks. Computational estimates of pitch strength, amplitude modulation filterbank output, and sharpness were obtained through custom-designed MATLAB algorithms. Perceived roughness and strain were significantly higher in children with SGVS and MVS compared to children with GVS. Among the computational measures, only the modulation filterbank output resulted in significant differences among vibratory sources; a posthoc test revealed that children with SGVS had greater amplitude modulation than children with GVS, as expected from their rougher voice quality. The results indicate that the output of an auditory amplitude modulation filterbank model may capture characteristics of SGVS that are strongly related to the rough voice quality.

Improving the sensitivity of early rub detection in rotating machines with an adaptive orthogonal filter

The early detection of rotor-casing rub helps to prevent permanent damage or catastrophic failure, saves maintenance costs and decreases the downtime of rotating machines. In aeroderivative gas turbines, rub detection is possible only with the use of casing sensors such as accelerometers. In previous works, the Wavelet Synchrosqueezed Transform (WSST) was proposed for very early detection of rub in aeroderivative gas turbines with accelerometer signals. In a quest for even better detection performances, this manuscript aims to design adaptive orthogonal wavelet filter banks for optimal rub feature separation from accelerometer signals. The outputs of the adaptive filter bank show improved feature separation between the no-rub and the rub regimes within a validation set from an experimental machine. Ultimately, it is demonstrated that the combination of the WSST and the adaptive orthogonal filter outperforms other methodologies for the early detection of rotor-casing rub with accelerometer signals, also when white noise is present in the data.

Binaural Model for Sound Image Width Estimation for Audio Listening Environment in Car Cabin

This paper proposes a binaural model for estimating the sound image width of the audio-listening environment in a car cabin. Based on a binaural model for the sound image azimuth estimation proposed in a previous study, a new feature was extracted and expanded into a model capable of estimating the sound image width. To improve the estimation accuracy of this model, the width of the sound image azimuth estimation distribution was extracted from the attack region of the gammachirp auditory filter bank outputs, and a multiple nonlinear regression model that uses the feature and sound image azimuth estimate value was constructed and examined. As the results indicate, the proposed method achieves a significantly higher accuracy.

Band importance for speech-in-speech recognition.

Predicting masked speech perception typically relies on estimates of the spectral distribution of cues supporting recognition. Current methods for estimating band importance for speech-in-noise use filtered stimuli. These methods are not appropriate for speech-in-speech because filtering can modify stimulus features affecting auditory stream segregation. Here, band importance is estimated by quantifying the relationship between speech recognition accuracy for full-spectrum speech and the target-to-masker ratio by channel at the output of an auditory filterbank. Preliminary results provide support for this approach and indicate that frequencies below 2 kHz may contribute more to speech recognition in two-talker speech than in speech-shaped noise.

Patient BBIs Monitoring System and Movement Analysis using Machine Learning Techniques and IoT

Wireless technology has been developed for different medical fields. The pulse monitoring system is used and in some private medical places only recorded and not utilized properly. However, accessing large medical equipment is inconvenient and expensive. In this, The proposed real-time heart rate monitoring system extract the heart beats measurements and that is stored and retrieved via Arduino microcontroller which connects the heart rate to the internet and sends results to the web server and notifies it to the specialist. Zero crossing in the resulting filter bank output signal represents heart rates that are extracted from BBI’s. The obtained heart rate is classified using the algorithm to denote the condition of the patient to notify it to specialist. Finally, the HRV features are calculated from the extracted BBI’s. If any risk found, it sends the emergency alert notification to the specialist. This system also monitors the patient activity in case of any emergency of patient using the IR sensor connected to the Arduino micro controller which alerts the controller or the specialist by a buzzer. The proposed framework isn't just utilized by the trained professional yet additionally by different clients (User agreeable). It offers speed and exactness in outcomes and with negligible expense. This technique is planned a Heart Rate Monitor System utilizing Arduino and Heartbeat Sensor. The exhibition results are contrasted and ECG signal addressed by oscilloscope and manual heartbeat estimation of heartbeat rate, which gives great arrangement.

Joint Channel Estimation and Equalization Using New AFB Output Signal Models for FBMC/OQAM Systems

Filter bank multicarrier with offset quadrature amplitude modulation (FBMC/OQAM) has been studied as an alternative waveform for post-orthogonal frequency division multiplexing (OFDM) wireless communications. However, due to the intrinsic imaginary inter-carrier/inter-symbol interference of FBMC/OQAM signals, accurate channel estimation and equalization remain open issues for FBMC/OQAM systems in dispersive channels. In this article, two new signal models of the analysis filter bank (AFB) output are first derived in the absence of any channel assumption and in the presence of a relaxed channel assumption, respectively. Based on the new-built models, initial channel estimation with least squares (LS) (IniCE-LS) is presented and its output is used as the initial condition of recursive weighted LS (RWLS) estimation; moreover, non-iterative and iterative channel equalization combined with RWLS channel estimation (i.e., N-ICE-RWLS and ICE-RWLS) are proposed by using the real-time reconstructed FBMC/OQAM data symbols as pilots. The proposed N-ICE-RWLS and ICE-RWLS schemes not only allow to improve channel estimation and equalization performance, but also achieve high transmission efficiency. Considering the requirements of practical application, we further develop the low-complexity implementation methods of N-ICE-RWLS and ICE-RWLS. All these schemes are validated by analyzing the channel estimation accuracy and bit error rate (BER) over various frequency selective channels.

The Hearing-Aid Speech Perception Index (HASPI) Version 2

This paper presents a revised version of the Hearing-Aid Speech Perception Index (HASPI). The index is based on a model of the auditory periphery that incorporates changes due to hearing loss and is valid for both normal-hearing and hearing-impaired listeners. It is an intrusive metric that compares the time-frequency envelope and temporal fine structure (TFS) of a degraded signal to an unprocessed reference. The first modification to HASPI is an extension to the range of envelope modulation rates considered in the metric. HASPI applies a lowpass filter to the time-frequency envelope modulation, and in the new version this single filter is replaced by a modulation filterbank. The temporal fine structure (TFS) analysis in the original version of HASPI is replaced by the filterbank outputs at higher modulation rates that represent auditory roughness and periodicity. The second modification is replacing the parametric model combining envelope and TFS measurements used in the original version with an ensemble of neural networks. The improved version of HASPI is compared to the original version for datasets from five experiments that encompass noise and nonlinear distortion, frequency compression, ideal binary mask noise suppression, speech modified using a noise vocoder, and speech in reverberation. The new version of HASPI is shown to have a statistically-significant reduction in RMS error compared to the original version for most of the data considered, and to be significantly more accurate for speech in reverberation.

Multi-Tap Channel Estimation for Preamble-Based FBMC/OQAM Systems

The intrinsic imaginary inter-carrier/inter-symbol interference in filter bank multicarrier with offset quadrature amplitude modulation (FBMC/OQAM) systems makes accurate channel estimation remain an open issue, which motivates our work in this article. In order to improve the channel estimation performance for preamble-based FBMC/OQAM systems, a recursive least squares (RLS)-based multi-tap channel estimation (RLS-MTCE) scheme that can effectively reduce the influence of noise is proposed by adopting the widely used signal model of the analysis filter bank (AFB) output. On the other hand, by utilizing the new-built signal model of the AFB output, the authors develop least squares (LS)-based and minimum mean square error (MMSE)-based multi-tap channel estimation schemes, denoted by LS-MTCE and MMSE-MTCE, respectively, to accurately exploit the interference to improve the channel estimation performance. The complexity and performance analysis results show that, compared with the existing schemes, the proposed RLS-MTCE, LS-MTCE and MMSE-MTCE counterparts generally achieve a lower estimation error at the cost of an acceptable increase in computational complexity.

Sidelobe Leakage Reduction in Random Phase Diversity Radar Using Coherent CLEAN

Random pulse-to-pulse phase diversity provides the means to suppress range-ambiguous interference in medium-to-high pulse repetition frequency radar applications. However, Doppler filters used in random phase diversity processing tend to suffer from high average sidelobe levels. This paper shows how this problem can be addressed using Coherent CLEAN, a deconvolution algorithm originally developed to address the problem of sidelobe interference in sparse antenna arrays. In the context of random phase diversity, Coherent CLEAN can remove sidelobe leakage from the output of a random phase diversity Doppler filter bank or be applied more directly to disentangle returns from multiple range ambiguities. The algorithm is proposed as an effective and computationally efficient feasible alternative to iterative filter refinement or “worst-case” filter shaping for suppression of unexpected returns. Multiple modes of applying Coherent CLEAN are described and a new variation of the algorithm, simultaneous Coherent CLEAN, is introduced.

High-Accuracy Real-Time Monitoring of Heart Rate Variability Using 24 GHz Continuous-Wave Doppler Radar

This paper presents a novel algorithm for the estimation of heart rate variability (HRV) features using 24-GHz continuous-wave Doppler radar with quadrature architecture. The proposed algorithm combines frequency and time domain analysis for high-accuracy estimation of beat-to-beat intervals (BBIs). Initially, band pass filtered in-phase (I) and quadrature (Q) radar components are fused into a single combined signal that contains information on the heartbeats. Its frequency domain analysis is used for coarse heart rate estimation. At the same time, the combined signal is processed using a filter bank containing narrowband band pass filters with different center frequencies. One of the band pass filter outputs is selected as the valid output based on the coarse heart rate estimation. Zero crossings in the resulting filter bank output signal represent heartbeats that are used to extract the BBIs. Finally, four HRV features are calculated from the BBIs. The algorithm is tested on real data obtained from recordings on ten human subjects. The mean relative error of extracted BBIs compared to electrocardiogram (ECG) measurement is in the 1.02–2.07% range. Furthermore, two time-domain and two frequency domain HRV features were calculated from the BBIs. The obtained results show a high level of agreement between radar-extracted and ECG-extracted HRV features. Low computation complexity makes this algorithm suitable for real-time monitoring.

The Factor Analysis of Speech: Limitations and Opportunities for Cochlear Implants

Current spread is known to limit the number of independent spectral channels in cochlear implants. The outcome of an experiment employing cochlear implant simulations indicated that current spread is not the only limitation on the benefit of increasing the number of electrodes: for both sentences and digit triplets, improvements in speech reception threshold slowed markedly once more than seven electrodes/processing channels were simulated. Factor analysis of speech envelopes from the output of an auditory filterbank confirmed that speech contains 6-8 independent sources of information, causing finer spectral filtering to produce redundant information in adjacent channels. It is possible that factor analysis can be used to refine the frequency maps used in cochlear implants in order to minimise the effects of current spread.

Design of high performance QRS complex detector for wearable healthcare devices using biorthogonal spline wavelet transform

A high performance QRS complex detector applicable for wearable healthcare devices is proposed in the present work. Since, higher SNR results in better detection accuracy and lesser number of coefficients reduces the hardware resources as well as power dissipation during on chip implementation. Biorthogonal spline wavelet transform is chosen for the proposed detector as it has high signal to noise ratio (SNR) and uses only four coefficients for decomposition. In the proposed approach, a Biorthogonal wavelet filter bank with fourth level decomposition is first used to separate the different frequency components and then a fourth level wavelet filter bank is used to get the denoised electrocardiogram (ECG) signals. Wavelet filter bank outputs are multiplied and soft threshold method is applied to get the QRS complex peaks by the QRS complex peak detector block. Add and shift multiplier used in the earlier designs has been replaced by a Booth multiplier in our approach to achieve the higher performance. Booth multiplier and QRS complex peak detector blocks have been designed for low hardware complexity, high performance and accurate detection of the QRS complex peaks. Time interval between the consecutive QRS peaks is calculated using the R-R peak time calculator block and the heart rate (HR) by the HR calculator block. Heart Rate Variability (HRV) and arrhythmia are detected based on these heart rate calculations. Proposed design has been tested for its robustness on multiple datasets (namely, MIT-BIH arrhythmia, MIT-BIH noise stress test, and MIT-BIH atrial fibrillation databases). Sensitivity of 99.31%, positive predictivity of 99.19% and the Detection Error Rate (DER) of 1.49% shown by the proposed design makes it preferable for QRS complex detectors used in wearable healthcare devices.

Machine listening in combination with microphone arrays for noise source localization and identification

In a recent project, a large microphone array system has been created to localize and quantify noise sources in an Intensive Care Unit (ICU). In the current state, the output of the system is the location and level of the most dominant noise sources, which is also presented in real-time to the nursing staff. However, both staff as well as patients have expressed the need for information about the types of noise sources. This additional source identification can also help to find means of reducing the overall noise level in the ICU. To accomplish the source identification, the approach of machine listening with a deep neural network is chosen. A feed-forward pattern recognition network is considered in this work. However, it is not clear which types of features are best suited for the given application. This contribution thus examines the problem from a practical point of view, comparing different features including those related to sound perception, such as specific loudness, Mel-frequency cepstral coefficients, as well as the output of a gamma-tone filter bank. Additionally, the concept of time-delay networks is tested to see whether a better classification of the signals can be achieved by including their time history.

Robust detection and segmentation of cell nuclei in biomedical images based on a computational topology framework.

The segmentation of cell nuclei is an important step towards the automated analysis of histological images. The presence of a large number of nuclei in whole-slide images necessitates methods that are computationally tractable in addition to being effective. In this work, a method is developed for the robust segmentation of cell nuclei in histological images based on the principles of persistent homology. More specifically, an abstract simplicial homology approach for image segmentation is established. Essentially, the approach deals with the persistence of disconnected sets in the image, thus identifying salient regions that express patterns of persistence. By introducing an image representation based on topological features, the task of segmentation is less dependent on variations of color or texture. This results in a novel approach that generalizes well and provides stable performance. The method conceptualizes regions of interest (cell nuclei) pertinent to their topological features in a successful manner. The time cost of the proposed approach is lower-bounded by an almost linear behavior and upper-bounded by O(n2) in a worst-case scenario. Time complexity matches a quasilinear behavior which is O(n1+ɛ) for ε < 1. Images acquired from histological sections of liver tissue are used as a case study to demonstrate the effectiveness of the approach. The histological landscape consists of hepatocytes and non-parenchymal cells. The accuracy of the proposed methodology is verified against an automated workflow created by the output of a conventional filter bank (validated by experts) and the supervised training of a random forest classifier. The results are obtained on a per-object basis. The proposed workflow successfully detected both hepatocyte and non-parenchymal cell nuclei with an accuracy of 84.6%, and hepatocyte cell nuclei only with an accuracy of 86.2%. A public histological dataset with supplied ground-truth data is also used for evaluating the performance of the proposed approach (accuracy: 94.5%). Further validations are carried out with a publicly available dataset and ground-truth data from the Gland Segmentation in Colon Histology Images Challenge (GlaS) contest. The proposed method is useful for obtaining unsupervised robust initial segmentations that can be further integrated in image/data processing and management pipelines. The development of a fully automated system supporting a human expert provides tangible benefits in the context of clinical decision-making.

Simple Feedforward Carrier Phase Estimation for Optical FBMC/OQAM Systems

Filter-bank multi-carrier based on offset quadrature amplitude modulation (OQAM) has recently been proposed to improve the spectral efficiency of optical fiber communication systems. Unfortunately, it is prone to the laser phase noise (PN) that results in significant inter-subcarrier interference due to the loss of OQAM orthogonality. Based on the analysis of the OQAM constellation obtained at the output of the analysis filter bank, we propose a modified blind phase search algorithm working based on the observation of all subcarriers simultaneously. It efficiently compensates for the PN and incurs a low number of multiplications. The proposed method is analytically derived and numerically validated with 4-, 16-, and 64-OQAM modulations. The impact of the number of subcarriers used per optical channel is also investigated. The results show that up to 256 subcarriers can be utilized with negligible performance degradation. Moreover, the method provides a tolerated linewidth and symbol duration product of $1.4\times 10^{\mathrm {-4}}$ , $2.5\times 10^{\mathrm {\mathbf {-5}}}$ , and $6.2\times 10^{\mathrm {-6}}$ for respective 4-, 16-, and 64-OQAM modulations to limit the SNR penalty to 1 dB at a BER of $10^{\mathrm {-3}}$ .

Analysis of acoustic features for speech intelligibility prediction models analysis of acoustic features for speech intelligibility prediction models

We have developed a new model to predict speech intelligibility of synthetic sounds processed by nonlinear speech enhancement algorithms. The model involves two recent auditory models: the dynamic compressive gammachirp (dcGC) auditory filterbank and the speech envelope power spectrum model (sEPSM). The dcGC-sEPSM was compared with commonly used prediction models based on perceptual intelligibility scores of speech sounds enhanced by classic spectral subtraction and state-of-the-art Wiener filtering. As a result, the dcGC-sEPSM predicted the scores better than the coherence SII (CSII), the short-time objective intelligibility (STOI), and the original sEPSM using the gammatone filterbank. There was, however, still inconsistency between the prediction and data. In this work, we show the analysis of acoustic features used in the prediction models. The CSII calculates the magnitude-squared coherence between clean and processed spectra to derive a signal-to-distortion ratio. The STOI calculates the correlation coefficients between the short-time frame vectors of clean and degraded sound at the output of the one-third octave filterbank. The sEPSM calculates the signal-to-noise ratio in the envelope modulation domain at the output of the auditory filterbank. We summarize the methods and discuss desirable features that improve speech intelligibility predictions.

Drum sound onset detection based on class separation using deep neural network

This study reports an onset detection method for the snare drum sounds for music signals constructed with plural instruments. We propose a method which detects an onset with a class separation using a deep neural network (DNN). To train the network, frames which include the waveform of the snare drum were extracted from music signal, and the log mel-filter bank channel outputs were calculated for each frame. The DNN was trained by using these feature parameters. The inputs were the calculated filter bank outputs, and the teaching signals were binary information whether to include the waveform of a snare drum or not. To detect onsets, candidate frames of the onset time were found based on changing in the power of the target music signal, and feature values were extracted from the candidate frames. Obtained values were used as the input data for the trained DNN, and detection of whether or not to include a waveform of a snare drum can be achieved. Experiments were conducted for 52 popular songs included in the music database, and an average F-measure was 0.73.

Whisper to normal speech conversion using pitch estimated from spectrum

We can perceive pitch in whispered speech, although fundamental frequency (F0) does not exist physically or phonetically due to the lack of vocal-fold vibration. This study was carried out to determine how people generate such an unvoiced pitch. We conducted experiments in which speakers uttered five whispered Japanese vowels in accordance with the pitch of a guide pure tone. From the results, we derived a multiple regression function to convert the outputs of a mel-scaled filter bank of whispered speech into the perceived pitch value. Next, using this estimated pitch value as F0, we constructed a system for conversion of whispered speech to normal speech. Since the pitch varies with time according to the spectral shape, it was expected that the pitch accent would be kept by this conversion. Indeed, auditory experiments demonstrated that the correctly perceived rate of Japanese word accent was increased from 55.5% to 72.0% compared with that when a constant F0 was used.

Improving speaker verification performance against long-term speaker variability

Speaker verification performance degrades when input speech is tested in different sessions over a long period of time chronologically. Common ways to alleviate the long-term impact on performance degradation are enrollment data augmentation, speaker model adaptation, and adapted verification thresholds. From a point of view in features of a pattern recognition system, robust features that are speaker-specific, and invariant with time and acoustic environments are preferred to deal with this long-term variability. In this paper, with a newly created speech database, CSLT-Chronos, specially collected to reflect the long-term speaker variability, we investigate the issues in the frequency domain by emphasizing higher discrimination for speaker-specific information and lower sensitivity to time-related, session-specific information. F-ratio is employed as a criterion to determine the figure of merit to judge the above two sets of information, and to find a compromise between them. Inspired by the feature extraction procedure of the traditional MFCC calculation, two emphasis strategies are explored when generating modified acoustic features, the pre-filtering frequency warping and the post-filtering filter-bank outputs weighting are used for speaker verification. Experiments show that the two proposed features outperformed the traditional MFCC on CSLT-Chronos. The proposed approach is also studied by using the NIST SRE 2008 database in a state-of-the-art, i-vector based architecture. Experimental results demonstrate the advantage of proposed features over MFCC in LDA and PLDA based i-vector systems.

Mean Hilbert envelope coefficients (MHEC) for robust speaker and language identification

Adverse noisy conditions pose great challenges to automatic speech applications including speaker and language identification (SID and LID), where mel-frequency cepstral coefficients (MFCC) are the most commonly adopted acoustic features. Although systems trained using MFCCs provide competitive performance under matched conditions, it is well-known that such systems are susceptible to acoustic mismatch between training and test conditions due to noise and channel degradations. Motivated by this fact, this study proposes an alternative noise-robust acoustic feature front-end that is capable of capturing speaker identity as well as language structure/content conveyed in the speech signal. Specifically, a feature extraction procedure inspired by the human auditory processing is proposed. The proposed feature is based on the Hilbert envelope of Gammatone filterbank outputs that represent the envelope of the auditory nerve response. The subband amplitude modulations, which are captured through smoothed Hilbert envelopes (a.k.a. temporal envelopes), carry useful acoustic information and have been shown to be robust to signal degradations. Effectiveness of the proposed front-end, which is entitled mean Hilbert envelope coefficients (MHEC), is evaluated in the context of SID and LID tasks using degraded speech material from the DARPA Robust Automatic Transcription of Speech (RATS) program. In addition, we investigate the impact of the dynamic range compression stage in the MHEC feature extraction process on performance using logarithmic and power-law nonlinearities. Experimental results indicate that: (i) the MHEC feature is highly effective and performs favorably compared to other conventional and state-of-the-art front-ends, and (ii) the power-law non-linearity consistently yields the best performance across different conditions for both SID and LID tasks.