Background Noise

A novel method of parametric active contours with geometric shape prior is presented in this paper. The main idea of the method consists in minimizing an energy function that includes additional information on a shape reference called a prototype. Shape prior introduced through a similarity measurement between evolving contour and Procrustes mean shape of desired object. This similarity measurement is full Procrustes distance between these two contours that is invariant with respect to similarity transformations (translation, scaling, and rotation). This extra shape knowledge enhances the model robustness to noise, occlusion and complex background. We also use gradient direction information in addition to gradient magnitude for more robustness to complex background. In this paper we introduce one important application of this new snake "3D object tracking". We obtain promising results for 3D object tracking which show the robustness of our method against noise, complex background, similarity transformations, occlusion, and changing viewpoint of 3D object.

This paper studies a novel decomposition technique, suitable for blind separation of linear mixtures of signals comprising finite-length symbols. The observed symbols are first modeled as channel responses in a multiple-input-multiple-output (MIMO) model, while the channel inputs are conceptually considered sparse positive pulse trains carrying the information about the symbol arising times. Our decomposition approach compensates channel responses and aims at reconstructing the input pulse trains directly. The algorithm is derived first for the overdetermined noiseless MIMO case. A generalized scheme is then provided for the underdetermined mixtures in noisy environments. Although blind, the proposed technique approaches Bayesian optimal linear minimum mean square error estimator and is, hence, significantly noise resistant. The results of simulation tests prove it can be applied to considerably underdetermined convolutive mixtures and even to the mixtures of moderately correlated input pulse trains, with their cross-correlation up to 10% of its maximum possible value.

This study examined the regulation of speech volume in hypophonic subjects with Parkinson's disease (PD) and age-and gender-matched controls. The ®rst two experiments investigated the ability of subjects with PD to automatically regulate speech volume in response to two types of implicit cue: (i) background noise (BGN) and (ii) instantaneous auditory feedback (IAF). Control subjects demonstrated the Lombard eect by automatically speaking louder when competing against increasing levels of background noise. They also showed the reverse eect, decreasing speech volume when increasing levels of facilitative instantaneous auditory feedback were provided. Subjects with PD demonstrated decreased overall speech volume; they were less able than controls to appropriately increase volume as background noise increased, and to decrease volume as IAF increased. Thus, subjects with PD demonstrated over-constancy of speech volume and failed to respond to the implicit cues integral to volumetric scaling. A further experiment (3) was carried out to contrast the regulation of volume in response to implicit cue with an explicit attention-driven cue (i.e. instructions regarding volume level). As in Experiments 1 and 2, subjects with PD exhibited reduced speech volume. Under explicit volume instructions, the ability of subjects with PD to regulate volume was normalised. These ®ndings suggest that subjects with PD have the capacity to speak with normal volume provided they consciously attend to speaking loudly. In subjects with PD, overall speech volume was always lower than for control subjects, suggesting a reduction of cortical motor set in the articulatory system similar to that demonstrated by the reduced amplitude of limb movements (hypokinesia) in the motor system. #

This paper presents a novel voice activity detector (VAD) for improving speech detection robustness in noisy environments and the performance of speech recognition systems in real-time applications. The algorithm is based on a generalized complex Gaussian (GCG) observation model and defines an optimal likelihood ratio test (LRT) involving multiple and correlated observations (MCO) based on jointly Gaussian probability distribution functions (jGpdf). An extensive analysis of the proposed methodology for a low dimensional observation model demonstrates 1) the improved robustness of the proposed approach by means of a clear reduction of the classification error as the number of observations is increased, and 2) the tradeoff between the number of observations and the detection performance. The proposed strategy is also compared to different VAD methods including the G.729, AMR, and AFE standards, as well as other recently reported algorithms showing a sustained advantage in speech/nonspeech detection accuracy and speech recognition performance.

Digital feedback loops for d.c. SQUIDs using pulse-rate modulation are proposed and investigated by simulations and measurements. The design of a d.c. SQUID magnetometer yields a simulated energy resolution of ~-11 h (h = Planck's constant) where ~ 3 h corresponds to additional noise due to the feedback loop. Measurements at Josephson sampler prototype chips showed that the digital feedback loop is very stable and fast enough to obtain a flicker-free display on an oscilloscope. In contrast to conventional feedback loops an integration of sensor and feedback loop on one or several chips with Josephson junctions seems feasible.

Acoustic echo canceler and postfilter for residual echo suppression are two essential building blocks of a hands-free voice communication system. Based on the Kalman filter theory, we derive a simple and advanced algorithm for the optimum joint statistical adaptation of both filter coefficients in time-varying and noisy acoustic environments. The Kalman filter utilizes a stochastic state-space model of the acoustic echo path which is formulated entirely in the frequency-domain. The resulting adaptive algorithm is computationally efficient and inherently robust, i.e., the adaptation does not require additional regularization or control mechanisms.

This paper describes a method for enhancing speech corrupted by broadband noise. The method is based on the spectral noise subtraction method. The original method entails subtracting an estimate of the noise power spectrum from the speech power spectrum, setting negative differences to zero, recombining the new power spectrum with the original phase, and then reconstructing the time waveform. While this method reduces the broadband noise, it also usually introduces an annoying "musical noise". We have devised a method that eliminates this "musical noise" while further reducing the background noise. The method consists in subtracting an overestimate of the noise power spectrum, and preventing the resultant spectral components from going below a preset minimum level (spectral floor). The method can automatically adapt to a wide range of signal-to-noise ratios, as long as a reasonable estimate of the noise spectrum can be obtained. Extensive listening tests were performed to determine the quality and intelligibility of speech enhanced by our method. Listeners unanimously preferred the quality of the processed speech. Also, for an input signal-to-noise ratio of 5 dB, there was no loss of intelligibility associated with the enhancement technique.

A 24-dimensional model for the ‘harmonic content’ of pieces of music has proved to be remarkably robust in the retrieval of polyphonic queries from a database of polyphonic music in the presence of quite significant noise and errors in either query or database document. We have further found that higher-order (1st- to 3rd-order) models tend to work better for music retrieval than 0th-order ones owing to the richer context they capture. However, there is a serious performance cost due to the large size of such models and the present paper reports on some attempts to reduce dimensionality while retaining the general robustness of the method. We find that some simple reduced-dimensionality models, if their parameter settings are carefully chosen, do indeed perform almost as well as the full 24-dimensional versions. Furthermore, in terms of recall in the top 1000 documents retrieved, we find that a 6-dimensional 2nd-order model gives even better performance than the full model. This represents a potential 64-times reduction in model size and search-time, making it a suitable candidate for filtering a large database as the first stage of a two-stage retrieval system.

This work focuses on presenting a novel high efficiency channel encoding for transmitting acoustical signals in the hope of improving wireless underwater communication. Yet, water puts a damper on communication capacity, slowing down the signal propagation and creating background noise and echoes. For thus systems and methods for implementing a control channel, e.g., in underwater communication system, are presented below. Aspects of the channel structures used to implement the control channel described herein, can improve error detection capabilities, reduce decoding complexity, and increase transmission efficiency. In certain aspects, transmission efficiency can be increased through using CRC encoding. A circular trellis check and Viterbi decoding can also be used to increase efficiency and maintain error detection capabilities. Symbol Error Rate (SER) can be reduced in embodiments described herein over that of tail-biting convolutional coding with a CRC. Furthermore, error detection offered by Reed Solomon encoder can well compensate underwater noise. Additionally, the encoder packet size can be fixed in order to facilitate decoding and reducing receiver complexity.

For Motion Picture Special Effects, it is often necessary to take a source image of an acto& segment the actor from the unwanted background, and then composite over a new background. The standard approach requires the unwanted background to be a blue screen. While this technique is ...

Speech reception in noise is an especially difficult problem for listeners with hearing impairment as well as for users of cochlear implants (CIs). One likely cause of this is an inability to 'glimpse' a target talker in a fluctuating background, which has been linked to deficits in temporal fine-structure processing. A finestructure cue that has the potential to be beneficial for speech reception in noise is fundamental frequency (F0). A challenging problem, however, is delivering the cue to these individuals. The benefits to speech intelligibility of F0 for both listeners with hearing impairment and users of CIs are reviewed, as well as various methods of delivering F0 to these listeners.

For Motion Picture Special Effects, it is often necessary to take a source image of an actor, segment the actor from the unwanted background, and then composite over a new background. The standard approach requires the unwanted background to be a blue screen. While this technique is capable of handling areas where the foreground blends into the background, the physical requirements present many practical problems. This paper presents an algorithm that requires minimal human interaction to segment Motion Picture resolution images and image sequences. We show that it can be used not only to segment badly lit or noisy bluescreen images, but also to segment actors where the background is more varied.

Seven Triticum aestivum (cv. Moisson)-Aegilops ventricosa addition lines and four VPM-1 lines were studied by C-banding, and compared with the parental common wheat cultivars Marne-Desprez (hereafter Marne), Moisson, and A. ventricosa lines 10 and 11. All of the VPM-1 lines had similar C-banding patterns and carried the same major 5B:7B translocation as the parental Marne cultivar. According to the C-banding analysis, the VPM-1 lines carry a complete 7D(7D v ) chromosome substitution and a translocation involving the 5D and 5D v chromosomes. However, the translocation of the 2N v /6N v chromosome of A. ventricosa to the short arm of the 2A chromosome of wheat that had been identiWed in an earlier study using molecular analysis (Bonhomme A, Gale MD, Koebner RMD, Nicolas P, Jahier J, Bernard M in Theor

A multi-element graphite furnace atomic absorption spectrometry (GFAAS) method was elaborated for the simultaneous determination of As, Cd, Cu, and Pb in wine samples of various sugar contents using the transversally heated graphite atomizer (THGA) with end-capped tubes and integrated graphite platforms (IGPs). For comparative GFAAS analyses, direct injection (i.e., dispensing the sample onto the IGP) and digestion-based (i.e., adding oxidizing agents, such as HNO 3 and/or H 2 O 2 to the sample solutions) methods were optimized with the application of chemical modifiers.

In radar signal detection, the problem is to automatically detect a target in a nonstationary noise and clutter background while maintaining a constant probability of false alarm. Classical detection using a matched filter receiver and a fixed threshold is not applicable due to the nonstationary nature of the background noise. Therefore, adaptive threshold techniques are needed to maintain a constant false-alarm rate (CFAR). One approach to adaptive detection in nonstationary noise and clutter background is to compare the processed target signal to an adaptive threshold. In the cell-averaging CFAR processing, an estimate of the background noise from the leading and the lagging reference windows is used to set the adaptive threshold. A threshold multiplier (or scaling factor) is used to scale the threshold to achieve the desired probability of false alarm. In the first part of this report, we have proposed two modified cell-averaging detectors for multiple target situations. The firs...

Man machine interaction requires an acoustic interface for providing full duplex hands-free communication. The transfer-function generalized sidelobe canceller (TF-GSC) is an adaptive beamformer suitable for enhancing a speech signal received by an array of microphones in a noisy and reverberant environment. When an echo signal is also present in the microphone output signals, cascade schemes of acoustic echo cancellation and TF-GSC can be employed for suppressing both interferences. However, the performances obtainable by cascade schemes are generally insufficient. An acoustic echo canceller (AEC) that precedes the adaptive beamformer suffers from the noise component at its input. Acoustic echo cancellation following the adaptive beamformer lacks robustness due to time variations in the echo path affecting beamformer adaptation. In this paper, we introduce an echo transfer-function generalized sidelobe canceller (ETF-GSC), which combines the TF-GSC with an acoustic echo canceller. The proposed scheme consists of a primary TF-GSC for dealing with the noise interferences, and a secondary modified TF-GSC for dealing with the echo cancellation. The secondary TF-GSC includes an echo canceller embedded within a replica of the primary TF-GSC components. We show that using this structure, the problems encountered in the cascade schemes can be appropriately avoided. Experimental results demonstrate improved performance of the ETF-GSC compared to cascade schemes in noisy and reverberant environments.

We have recently extended the capabilities of BBN's large-vocabulary discrete-utterance speech recognition system (BYBLOS) to operate on raw audio recordings of radio news programming. The recordings are given to the system as large monolithic ...

Traditional noise-suppression algorithms have been shown to improve speech quality, but not speech intelligibility. Motivated by prior intelligibility studies of speech synthesized using the ideal binary mask, an algorithm is proposed that decomposes the input signal into time-frequency ͑T-F͒ units and makes binary decisions, based on a Bayesian classifier, as to whether each T-F unit is dominated by the target or the masker. Speech corrupted at low signal-to-noise ratio ͑SNR͒ levels ͑−5 and 0 dB͒ using different types of maskers is synthesized by this algorithm and presented to normal-hearing listeners for identification. Results indicated substantial improvements in intelligibility ͑over 60% points in −5 dB babble͒ over that attained by human listeners with unprocessed stimuli. The findings from this study suggest that algorithms that can estimate reliably the SNR in each T-F unit can improve speech intelligibility.

A bstract-We consider the optical direct-detection channel and show how simple trellis coded modulation (TCM) can be used to improve performance or increase throughput (in bits per second) without bandwidth expansion or performance loss. In fact, a modest performance gain can be achieved. Although the approach can be used with other signal constellations, we concentrate on signals derived from the pulse-position modulation (PPM) format by allowing overlap. Theoretical motivation for using this signal set, known as overlapping PPM (OPPM), was recently given by Bar David et al. who showed a capacity gain when overlap is introduced.

In this paper we introduce a new adaptive array beamformer able to work well even when the desired signal and the interference are coherent. The present adaptive beamformers fail to operate in these cases. The results of simulations support the theoretical predictions.

The classification of acoustic events in indoor environments is an important task for many practical applications in smart environments. In this paper a novel approach for classifying acoustic events that is based on a Bag-of-Features approach is proposed. Mel and gammatone frequency cepstral coefficients that originate from psychoacoustic models are used as input features for the Bag-of representation. Rather than using a prior classification or segmentation step to eliminate silence and background noise, Bag-of-Features representations are learned for a background class. Supervised learning of codebooks and temporal coding are shown to improve the recognition rates. Three different databases are used for the experiments: the CLEAR sound event dataset, the D-CASE event dataset and a new set of smart room recordings.

The performance of present-day automatic speech recognition (ASR) systems is seriously compromised by levels of acoustic interference (such as additive noise and room reverberation) representative of real-world speaking conditions. Studies on the perception of speech by human listeners suggest that recognizer robustness might be improved by focusing on temporal structure in the speech signal that appears as low-frequency (below 16 Hz) amplitude modulations in subband channels following critical-band frequency analysis. A speech representation that emphasizes this temporal structure, the``modulation spectrogram'', has been developed. Visual displays of speech produced with the modulation spectrogram are relatively stable in the presence of high levels of background noise and reverberation. Using the modulation spectrogram as a front end for ASR provides a signi®cant improvement in performance on highly reverberant speech. When the modulation spectrogram is used in combination with log-RASTA-PLP (log RelAtive SpecTrAl Perceptual Linear Predictive analysis) performance over a range of noisy and reverberant conditions is signi®cantly improved, suggesting that the use of multiple representations is another promising method for improving the robustness of ASR systems.

Sound absorption and optical transparency are among the most useful properties of noise barriers. While the latter is required to reduce visual impact and for aesthetical reasons, the former is required whenever conditions of multiple reflections and presence of close, high receivers occur. The technical feasibility of a transparent, sound-absorbing panel for outdoor antinoise devices is investigated in this paper. An analysis of acoustical performance of multiple perforated plates is performed employing an existing theory for microperforated absorbers under normal incidence and diffused sound field. An optimization of the geometrical parameters is carried out on the basis of the European classification criteria of noise barriers for roadways. An optimized three-layer configuration can achieve sound-absorption properties similar to nontransparent products with only a limited loss of visual transparency and appropriate mechanical strength. Experimental data obtained with an impedance tube on small test samples made of transparent polycarbonate and in a reverberation room on full-scale prototypes are reported, showing a rather good agreement with the theoretical predictions. The optical performance of a multilayered configuration is evaluated also.

In the present study a new approach for the chemometric background correction in on-line gradient LC-FTIR is introduced. For this purpose, the spectral changes of the elution mixture during gradient elution were analyzed applying 2D correlation spectroscopy. The fundamentals of the new background correction algorithm, based on polynomial fits calculated from a reference spectra matrix (Polyfit-RSM method) are explained. The Polyfit-RSM approach was applied on blank gradient runs as well as on LC-FTIR data obtained from the injection of a soft drink sample using acetonitrile:water as eluent. Results found were critically assessed and compared to those obtained by two previous background correction methods which are likewise based on the use of a reference spectra matrix (RSM). The Polyfit-RSM method provided lower noise levels throughout the whole spectral range than other alternative background correction methods, an excellent recovery of analyte spectra as well as chromatograms with a low noise level and also free from baseline shifts. A significant finding, which implies a major advantage for the practical applicability of the algorithm, is that the size of the RSMs can be reduced without affecting the accuracy of the correction method.

In the digital smoothing of data in AES and XPS two basic methods. Savitzky and Golay convolutional sm(n~thing and the optimal Wiener filter Fourier transform approach are popular. In the recent work of Piyakis and Sachet misleadingly biased statements are expressed about their relative merils. Here we delineate, as completely as we can. most of the basic principles which show an advantage of transparenc~ in the applicatkm of the Savitzky a.',t! Golay method compared with an advantage of speed for the Fourier Transform method. In terms of mathematical principles both methods can extract as much information as is available.

An improved hidden Markov model-based (HMMbased) speech enhancement system designed using the minimum mean square error principle is implemented and compared with a conventional spectral subtraction system. The improvements to the system are: 1) incorporation of mixture components in the HMM for noise in order to handle noise nonstationarity in a more flexible manner, 2) two efficient methods in the speech enhancement system design that make the system realtime implementable, and 3) an adaptation method to the noise type in order to accommodate a wide variety of noises expected under the enhancement system's operating environment. The results of the experiments designed to evaluate the performance of the HMM-based speech enhancement systems in comparison with spectral subtraction are reported. Three types of noise-white noise, simulated helicopter noise, and multitalker (cocktail party) noise-were used to corrupt the test speech signals. Both objective (global SNR) and subjective mean opinion score (MOS) evaluations demonstrate consistent superiority of the HMM-based enhancement systems that incorporate the innovations described in this paper over the conventional spectral subtraction method.

The North Atlantic right whale inhabits the coastal waters off the east coasts of the United States and Canada, areas characterized by high levels of shipping and fishing activities. Acoustic communication plays an important role in the social behavior of these whales and increases in low-frequency noise may be leading to changes in their calling behavior. This study characterizes the ambient noise levels, including both natural and anthropogenic sources, and right whale upcall parameters in three right whale habitat areas. Continuous recordings were made seasonally using autonomous bottom-mounted recorders in the Bay of Fundy, Canada ͑2004, 2005͒, Cape Cod Bay, ͑2005, 2006͒, and off the coast of Georgia ͑2004-2005, 2006-2007͒. Consistent interannual trends in noise parameters were found for each habitat area, with both the band level and spectrum level measurements higher in the Bay of Fundy than in the other areas. Measured call parameters varied between habitats and between years within the same habitat area, indicating that habitat area and noise levels alone are not sufficient to predict variability in call parameters. These results suggest that right whales may be responding to the peak frequency of noise, rather than the absolute noise level in their environment.

This paper addresses the problem of radar target detection in severely heterogeneous clutter environments. Speciÿcally, we present the performance of the normalized matched ÿlter test in a background of disturbance consisting of clutter having a covariance matrix with known structure and unknown scaling plus background white Gaussian noise. It is shown that when the clutter covariance matrix is low rank, the (LRNMF) test retains invariance with respect to the unknown scaling as well as the background noise level and has an approximately constant false alarm rate (CFAR). Performance of the test depends only upon the number of elements, the number of pulses processed in a coherent processing interval, and the rank of the clutter covariance matrix. Analytical expressions for calculating the false alarm and detection probabilities are presented. Performance of the method is shown to degrade with increasing clutter rank especially for low false alarm rates. An adaptive version of the test (LRNAMF) is developed and its performance is studied with simulated data from the KASSPER program. Results pertaining to sample support for subspace estimation, CFAR, and detection performance are presented. Target contamination of training data has a deleterious impact on the performance of the test. Therefore, a technique known as self-censoring reiterative fast maximum likelihood/adaptive power residue (SCRFML/APR) is developed to treat this problem and its performance is discussed. The SCRFML/APR method is used to estimate the unknown covariance matrix in the presence of outliers. This covariance matrix estimate can then be used in the LRNAMF or any other eigen-based adaptive processing technique.

Rolling-element bearing vibrations are random cyclostationary. This property is so symptomatic when an incipient fault develops that it can be exploited for diagnostics. This paper discusses which cyclic spectral tools should be considered for that purpose. Specifically, it demonstrates the optimality of the cyclic coherence, which can not only evidence the presence of a fault in high levels of background noise, but can also return a relative measure of its severity. The estimation issue of the cyclic coherence is addressed in detail, as well as its use in a statistical test, and sub-optimal simplifications. Eventually, it is shown that the familiar squared-envelope spectrum happens to be a special case of the cyclic coherence with very similar diagnostic capabilities.

The "noise" of chromatographic baselines has been investigated in regard to the detector, the nature and extent of filtering or smoothing. and the methodologies of qualitative and quantitative assessment: a11 in order to clarify the role such factors play in the determination and interconversion of some common types of detection firnits. This study scrutinizes baselines from the flame photometric detector in sjn~e-channel continuous and ten-channel multiplexing versions; it also examines baselines from flame ianization and electron-capture detectors. It makes we of finite impulse response and non-weighted moving-average digital smoothing, as well as three-pole analog filtering. Baseline fluctuations are quantified by the standard deviation derived from the common root-mean-square (RMS) calculation, or from the Iess common least-squares Gaussian fit; peak-to-peak noise (N,_,) is estimated by procedures including or excluding pre$umed outliers. Individual results are expressed as the ratio of NV_, measurement and RMS calculation performed on the same data set. A wide variety of such ratios are then assembled from different detectors, filters, and smoothing conditions. They prove conclusively that-contrary to common belief-the conversion factor between the two types of measurements does vary: usually between 4 and 10, but occasionally even farther. Consequently, the conversion factor between the corresponding two types of detection limits varies as well. The N,_,/RMS ratio depends largely on the detector-~~f~origin, its condition, and the extent to which noise has been filtered. In contrast, the nature and sophistication of the filter hardly matters: either for the N,_,iRMS ratio or for the practical detection limit. This is because the slow undulations characteristic of heavily filtered baselines represent-at least in the detectors we used-dampened fast noise rather than aboriginally slow noise. Corresponding computer simulations, based on amplitudinally random noise smoothed by stationary boxcar or non-weighted moving-average filters, produce results strikingly similar to actuat baselines. Simula& fast RMS noise correlates, as expected, with the square root (log-log slope = 112) of the filter's time constant. The corresponding slopes for experimetzrai noise are usuafly close to 112 as well. Most important&, though, the simuhred N_,lRMS ratio varies strongly with the extent of smoothing-thus mimicking and thereby explaining the behavior of the experimental ratio.

In MPEG-4, the boundary blocks of reference video objects are padded by replicating the boundary samples towards the exterior. This scheme does not make use ofthe trend of pixel value variation 0ft.n present near object boundaries in padding the exterior pixels ofthe reference video object. Rather, it results in a break ofthis trend at the video object boundary, resulting in higher prediction errors for those pixel locations that are within the shape of the current block, but are outside that of the predicted block. In this paper, we propose a novel shape distortion adaptive hybrid padding technique to minimise prediction errors and improve compression ef3ciency. We show that this technique results in boundary block coding gains of up to 9% as compared to that ofthe MPEG-4 scheme.

This paper presents a statistical measure for the identification of the presence, the location, and the calibration of the strength of singularity in a signal or in any of its derivatives in the presence of measurement noise without the requirement of a baseline using a wavelet based detection technique. For this proposed wavelet based detection of singularities present in a signal, the problem of false alarm and its significant reduction by use of multiple measurements is presented. The importance of the proposed measure on baseline and nonbaseline damage calibration has been discussed from the aspect of structural health monitoring. The findings in this paper can also be used for crosschecking of background noise level in an observed signal. The detection of the existence, location, and extent of an open crack from the first fundamental modeshape of a simply supported beam is presented as an example problem.

We report the first experimental demonstration of two-photon imaging with a pseudo-thermal source. Similarly to the case of entangled states, a two-photon Gaussian thin lens equation is observed, indicating EPR type correlation in position. We introduce the concepts of two-photon coherent and two-photon incoherent imaging. The differences between the entangled and the thermal cases are explained in terms of these concepts.

Identification functions of 20 initial and 20 final consonants were characterized in 9600 randomly sampled consonant-vowel-consonant ͑CVC͒ tokens presented in speech-spectrum noise. Because of differences in the response criteria for different consonants, signal detection measures were used to quantify identifiability. Consonant-specific baseline signal-to-noise ratios ͑SNRs͒ were adjusted to produce a dЈ of 2.20 for each consonant. Consonant identification was measured at baseline SNRs ͑B͒, at B − 6, and at B + 6 dB. Baseline SNRs varied by more than 40 dB for different consonants. Confusion analysis revealed that single-feature place-of-articulation errors predominated at the highest SNR, while combined-feature errors predominated at the lowest SNR. Most consonants were identified at lower SNRs in initial than final syllable position. Vowel nuclei ͑/Ä/, /i/, or /u/͒ significantly influenced the identifiability of 85% of consonants, with consistent vowel effects seen for consonant classes defined by manner, voicing, and place. Manner and voicing of initial and final consonants were processed independently, but place cues interacted: initial and final consonants differing in place of articulation were identified more accurately than those sharing the same place. Consonant identification in CVCs reveals contextual complexities in consonant processing.

Robust detection of speech embedded in real acoustic background noise is considered using an approach based on subband amplitude modulation spectral (AMS) features and trained discriminative classifiers. Performance is evaluated in particular for situations in which speech is embedded in acoustic backgrounds not presented during classifier training, and for signal-to-noise ratios (SNR) from-10 dB to 20 dB. The results show that (1) Generalization to novel background classes with AMS features yields better performance in 84% of investigated situations, corresponding to an SNR benefit of about 10 dB compared to mel-frequency cepstral coefficient (MFCC) features. (2) On known backgrounds, AMS and MFCCs achieve similar performance, with a small advantage for AMS in negative SNR regimes. (3) Standard voice activity detection (ITU G729.B) performs significantly worse than the classification-based approach.

The Australian Parliamentary Chambers present difficult circumstances for the reinforcement of speech, including an unusually wide dynamic range, large orator-to-microphone distances, high levels of uproar noise, and architectural and acoustical constraints. These circumstances were a challenge to designing a sound system that would provide high intelligibility with faithful spectral reproduction, wide angles of coverage, and a high acoustic gain and sound pressure level capability. Recognizing the interdependence of all factors, a holistic process was used to design and specify hardware. The solution included tapered line array loudspeakers and the use of original software. Commissioning the system involved numerical optimization of the gain structure and a novel technique that uses measurements of acoustic loop gain to predict equalizations that optimize the acoustic gain margin.

Multiaccess communications over a slot-synchronized collision channel without feedback (CCw/oFB) is considered. In the CCw/oFB the sources do not transmit siinultsiieously and continuously over the channel but according to protocol signals with duty factors smaller than or equal to one. This model was proposed by Massey and Mathys (11 for determining the loss due to sharing a common channel without timesharing. In their model the users share a common communication resource but the lack of a common time reference prevents operation in timesharing mode. Each user makes his transmissions at times determined by a protocol signal that is independent of the data to be sent, so that the capacity is independent of the source. This model expresses a worst-case assumption that each user has a nonempty queue of messages awaiting transmission. The slot-synchronized CCw/oFB is a block interference channel (BIC) [2] with two noise levels. The noise in interfered slots has infinite power (reflecting a worst case condition) while in noninterfered slots only white gaussian background noise is assumed to exist. In coded BIC's an interleaver is usually inserted for reducing the memory. Accordingly we have modified the CCw/oFB model to include interleavers, as depicted in Fig. . A complete loss of memory is achieved by ideal interleaving, i.e. by transmitting all code symbols of a codeword in different slots. However, the allowed delay may limit the interleaving span N to be of finite length, frequently even smaller than the codeword length (in letters). In such cases nonideal interleaving is obtained and the memory is not fully dispersed.

This study investigates masking effects occurring during speech comprehension in the presence of concurrent speech signals. We examined the differential effects of acoustic–phonetic and lexical content of 4- to 8-talker babble (natural speech) or babble-like noise (reversed speech) on word identification. Behavioral results show a monotonic decrease in speech comprehension rates with an increasing number of simultaneous talkers in the reversed condition. Similar results are obtained with natural speech except for the 4-talker babble situations. An original signal analysis is then proposed to evaluate the spectro-temporal saturation of composite multitalker babble. Results from this analysis show a monotonic increase in spectro-temporal saturation with an increasing number of simultaneous talkers, for both natural and reversed speech. This suggests that informational masking consists of at least acoustic–phonetic masking which is fairly similar in the reversed and natural conditions and lexical masking which is present only with natural babble. Both effects depend on the number of talkers in the background babble. In particular, results confirm that lexical masking occurs only when some words in the babble are detectable, i.e. for a low number of talkers, such as 4, and diminishes with more talkers. These results suggest that different levels of linguistic information can be extracted from background babble and cause different types of linguistic competition for target-word identification. The use of this paradigm by psycholinguists could be of primary interest in detailing the various information types competing during lexical access.

An efficient method for the enhancement of lung sounds (LS) and bowel sounds (BS), based on wavelet transform (WT), and fractal dimension (FD) analysis is presented in this paper. The proposed method combines multiresolution analysis with FD-based thresholding to compose a WT-FD filter, for enhanced separation of explosive LS (ELS) and BS (EBS) from the background noise. In particular, the WT-FD filter incorporates the WT-based multiresolution decomposition to initially decompose the recorded bioacoustic signal into approximation and detail space in the WT domain. Next, the FD of the derived WT coefficients is estimated within a sliding window and used to infer where the thresholding of the WT coefficients has to happen. This is achieved through a self-adjusted procedure that iteratively "peels" the estimated FD signal and isolates its peaks produced by the WT coefficients corresponding to ELS or EBS. In this way, two new signals are constructed containing the useful and the undesired WT coefficients, respectively. By applying WT-based multiresolution reconstruction to these two signals, a first version of the desired signal and the background noise is provided, accordingly. This procedure is repeated until a stopping criterion is met, finally resulting in efficient separation of the ELS or EBS from the background noise. The proposed WT-FD filter introduces an alternative way to the enhancement of bioacoustic signals, applicable to any separation problem involving nonstationary transient signals mixed with uncorrelated stationary background noise. The results from the application of the WT-FD filter to real bioacoustic data are presented and discussed in an accompanying paper.

Hands-free devices are often used in a noisy and reverberant environment. Therefore, the received microphone signal does not only contain the desired near-end speech signal but also interferences such as room reverberation that is caused by the near-end source, background noise and a far-end echo signal that results from the acoustic coupling between the loudspeaker and the microphone. These interferences degrade the fidelity and intelligibility of near-end speech. In the last two decades, postfilters have been developed that can be used in conjunction with a single microphone acoustic echo canceller to enhance the near-end speech. In previous works, spectral enhancement techniques have been used to suppress residual echo and background noise for single microphone acoustic echo cancellers. However, dereverberation of the near-end speech was not addressed in this context. Recently, practically feasible spectral enhancement techniques to suppress reverberation have emerged. In this paper, we derive a novel spectral variance estimator for the late reverberation of the near-end speech. Residual echo will be present at the output of the acoustic echo canceller when the acoustic echo path cannot be completely modeled by the adaptive filter. A spectral variance estimator for the so-called late residual echo that results from the deficient length of the adaptive filter is derived. Both estimators are based on a statistical reverberation model. The model parameters depend on the reverberation time of the room, which can be obtained using the estimated acoustic echo path. A novel postfilter is developed which suppresses late reverberation of the near-end speech, residual echo and background noise, and maintains a constant residual background noise level. Experimental results demonstrate the beneficial use of the developed system for reducing reverberation, residual echo, and background noise. ). P. C. W. Sommen is with the Signal Processing

The paper proposes a performance evaluation and comparison of G.729, AMR, and fuzzy voice activity detection (FVAD) algorithms. The comparison was made using objective, psychoacoustic, and subjective parameters. A highly varied speech database was also set up to evaluate the extent to which VADs depend on language, the signal-to-noise ratio (SNR), or the power level.