Echo Suppression Research Articles

A microphone array system with echo canceller

AbstractThis paper proposes a microphone array unified with an echo canceller that resolves the problem of deteriorated echo suppression performance due to variations of the echo path when a microphone array and an echo canceller are connected. In a teleconferencing system, high‐quality sound acquisition with noise suppression is desirable in addition to acoustic echo suppression. However, in a configuration in which a microphone array to suppress the noise and an echo canceller to suppress the echo are simply connected, the adaptation for the echo canceller cannot follow the echo path variations due to directivity control of the microphone array, so that there is a problem of deteriorated echo suppression performance. In the microphone array with an echo canceller proposed in this paper, the training of the microphone array and that of the echo canceller are performed in a unified way. In this way, the problem of deteriorated echo suppression performance due to echo path variations can be resolved. An evaluation experiment shows that the proposed method can perform stable noise suppression and echo cancellation. © 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 89(10): 23–32, 2006; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20090

Sound reinforcement system having an echo suppressor and loudspeaker beamformer

A sound reinforcement system (1) comprises several microphones (2), a microphone beamformer (5) coupled to the microphones (2), adaptive echo compensation (EC) means (4) coupled to the microphone beamformer (5) for generating an echo compensated microphone signal, and several loudspeakers (3) coupled to the adaptive EC means (4). The sound reinforcement system (1) further comprises an adaptive loudspeaker beamformer (11) coupled between the adaptive EC means (4) and the loudspeakers (3) for shaping the directional pattern of the loudspeakers (3). Advantageously the adaptive loudspeaker beamformer creates a beam pattern which is capable of creating a 'null' in the direction of speaker(s) such that howling is effectively prevented. The loudspeaker beamformer (11) may for example be a Weighted Sum Beamformer, a Delay and Sum Beamformer or a Filtered Sum Beamformer.

The absence of spatial echo suppression in the echolocating bats Megaderma lyra and Phyllostomus discolor

Acoustic orientation most often takes place in echoic environments. The mammalian auditory system shows a variety of specializations to suppress misleading spatial information mediated by echoes. Psychophysically these specializations are summarized as the precedence effect. This study investigates how echolocating bats deal with multiple reflections of their sonar emissions from different spatial positions. In a two-alternative, forced choice paradigm, a study was made of the extent to which the echolocating bats Megaderma lyra and Phyllostomus discolor spontaneously suppress the spatial information of a second reflection of their sonar emission. The delay between the first and the second reflection ranged between 0 and 12.8 ms. In general, M. lyra (five individuals) and P. discolor (two individuals) did not suppress the spatial information of the second reflection of their sonar emission, whatever the delay. Only one M. lyra showed significant suppression for delays between 0.8 and 3.2 ms. However, this suppression could not be confirmed in an exact repetition of the experiment. The current data indicate that although bats may be able to suppress the spatial information of a second reflection, this is not their default mode of auditory processing. The reason for this exceptional absence of spatial echo suppression may lie in the shorter time constants of cochlear processing in the ultrasonic frequency range and the strong influence of cognitive components associated with the ;precedence effect'.

Frequency-domain adaptive Kalman filter for acoustic echo control in hands-free telephones

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.

Residual echo filter for enhanced acoustic echo control

This paper presents a method for efficient control of an acoustic echo control (AEC) system. The control is achieved with the help of so-called residual echo filter structure. The residual echo filter can be exploited to provide an accurate estimate of the norm of the system mismatch vector (sometimes also called as an echo coupling factor), which is an essential element for providing advanced control means for the adaptive echo cancellation filters. In addition to the adaptive filter a sophisticated AEC systems usually include echo suppression means to complete and ensure sufficient echo cancellation performance. The residual echo filter may also be utilized to estimate the power spectral density of the residual echo, which is needed to implement the suppression filter. The use of the residual echo filter structure overcomes some weaknesses in the ordinary AEC controlling means and provides enhanced overall AEC performance.

Speech reinforcement system for car cabin communications

A speech reinforcement system is presented to improve communication between the front and the rear passengers in large motor vehicles. This type of communication can be difficult due to a number of factors, including distance between speakers, noise and lack of visual contact. The system described makes use of a set of microphones to pick up the speech of each passenger, then it amplifies these signals and plays them back to the cabin through the car audio loudspeaker system. The two main problems are noise amplification and electro-acoustic coupling between loudspeakers and microphones. To overcome these problems the system uses a set of acoustic echo cancellers, echo suppression filters and noise reduction stages. In this paper, the stability of a speech reinforcement system is studied. We propose a solution based on echo cancellers and residual echo suppression filters. The spectral estimation method for the power spectral density of the residual echo existing after the echo canceller is presented along with the derivation of the optimal residual echo suppression filter. Some results about the performance of the proposed system are also provided.

Suppressing acoustic echo in a spectral envelope space

Full-duplex hands-free telecommunication systems employ an acoustic echo canceler (AEC) to remove the undesired echoes that result from the coupling between a loudspeaker and a microphone. Traditionally, the removal is achieved by modeling the echo path impulse response with an adaptive finite impulse response (FIR) filter and subtracting an echo estimate from the microphone signal. It is not uncommon that an adaptive filter with a length of 50-300 ms needs to be considered, which makes an AEC highly computationally expensive. In this paper, we propose an echo suppression algorithm to eliminate the echo effect. Instead of identifying the echo path impulse response, the proposed method estimates the spectral envelope of the echo signal. The suppression is done by spectral modification-a technique originally proposed for noise reduction. It is shown that this new approach has several advantages over the traditional AEC. Properties of human auditory perception are considered, by estimating spectral envelopes according to the frequency selectivity of the auditory system, resulting in improved perceptual quality. A conventional AEC is often combined with a post-processor to reduce the residual echoes due to minor echo path changes. It is shown that the proposed algorithm is insensitive to such changes. Therefore, no post-processor is necessary. Furthermore, the new scheme is computationally much more efficient than a conventional AEC.

Auditory gating in schizophrenia: a pilot study of the precedence effect

Individuals with schizophrenia are known to exhibit diminished auditory gating, as manifested by prepulse inhibition of the startle response and suppression of the P50 evoked potential. Those observations indicate that schizophrenics have deficits of inhibitory auditory processing. The precedence effect is a phenomenon in which inhibitory processing in the auditory system aids the localization of sound sources and segregation of speech in complex auditory environments. We tested the hypothesis that schizophrenic subjects would exhibit deficits in psychophysical tests of the precedence effect. The performance of 12 male medicated schizophrenic subjects was compared with that of 12 male comparison subjects. Schizophrenics and comparison subjects clearly demonstrated all three aspects of the precedence effect that were evaluated: echo suppression, buildup of echo suppression, and breakdown of echo suppression. In none of these aspects did the performance of schizophrenic subjects differ significantly from that of the comparison subjects. These findings suggest that abnormalities of auditory spatial processing and speech segregation are unlikely to contribute to the clinical pathology of schizophrenia.

New SPUDT cell structures

The present paper describes single-phase unidirectional transducers (SPUDT) cells with all fingers wider than lambda/8 while maintaining the unidirectional effect. The first solution is related to a SPUDT consisting of lambda/4 and lambda/2 wide fingers arranged in two tracks. Each track has no significant unidirectional effect. Both tracks form a waveguide, and the waveguide coupling generates the interaction of the tracks. As a result of that interaction, a unidirectional effect arises as verified by experiment. This transducer type is called double-track (DT) SPUDT. A second solution is suggested that includes, in contrast to distributed acoustic reflection transducer (DART), electrode width control (EWC), and Hunsinger cells, SPUDT cell fingers with one and the same width only. Cell types with lambda/6, lambda/5, and lambda/3 wide fingers called uniform width electrode (UWE) cells are considered. One of these cell types, including exclusively lambda/5 wide fingers, is experimentally investigated and a unidirectional effect is found. Moreover, a filter example using the lambda/5 cell type has been designed for reducing SPUDT reflections. The echo suppression expected could be verified experimentally. No waveguide coupling is required for this cell type.

A detection-theoretic model of echo inhibition.

A detection-theoretic analysis of the auditory localization of dual-impulse stimuli is described, and a model for the processing of spatial cues in the echo pulse is developed. Although for over 50 years "echo suppression" has been the topic of intense theoretical and empirical study within the hearing sciences, only a rudimentary understanding of its mechanisms has emerged. In this article, psychometric functions and results from matching studies are used in developing a model that specifies the perceived position of the echo pulse as a normal deviate, with an expectation that is a logistic function of the echo delay and a variance that depends on interaural time difference. Loss of information in the echo event is quantified as a decline in the efficiency with which the binaural system receives information from the lag impulse.

Method and device for suppressing echo in a hands free device such as a telephone

An method and a device for echo suppression in a handsfree apparatus, particularly for a telephone is disclosed. With the aid of a control unit, control signals are generated from the signal of the far-end subscriber, the output signal of the loudspeaker-room-microphone system and the output signal of an adaptive filter used as reference, the control signals having an amplitude and a time response dependent on the speech mode. Using the control signals, the amplifications V E , V S of controllable amplifiers disposed in the receiving and transmitting paths are adapted to the speech mode, thus reducing the residual echo. The stepless transition of the amplifications from a quiescent value to a value attenuated with the factor and having an exponential time characteristic in response to a change of speech mode, provides an impression of natural speech.

Performance limits in subband beamforming

This paper analyzes subband beamforming schemes mainly aimed at speech enhancement and acoustic echo suppression applications such as hands-free telephony for both mobile and office environments, Internet telephony and video conferencing. Analytical descriptions of both causal finite-length and noncausal infinite-length subband microphone array structures are given. More specifically, this paper compares finite Wiener filter performance with the noncausal Wiener solution, giving a comprehensive theoretical suppression limit. It is shown that even short filters will yield a good approximation of the infinite solution, provided that the element spacing and temporal sampling is matched to the frequency band of interest. Typically, 10-20 FIR taps are sufficient in each subband.

Suppression of spurious echo in angle beam testing in automatic ultrasonic testing system for round bar

Suppression of spurious echo in angle beam testing in automatic ultrasonic testing system for round bar

Gap detection and location in the precedence effect

The nature of the precedence effect was investigated by introducing a gap into (1) both the leading and lagging sounds, (2) the lagging but not the leading sound, and (3) the leading but not the lagging sound. When a 50-ms gap was introduced into both sounds with an onset asynchrony equal to the delay between the leading and lagging sounds, the gap was perceived to occur only on the leading side as long as the delay between leading and lagging sounds did not exceed approximately 15 ms, even though the precedence effect itself broke down when the delay between the leading and lagging sounds exceeded approximately 9 ms. Gaps presented only in lagging sounds were always heard as occurring in the source position of the leading sound, but no gaps were perceived when the gap occurred only in the leading sound, rather, the listener heard a noise burst from the position of the lagging (suppressed) sound. The present results indicate that gaps in the lagging sound are perceived as belonging to the leading sound, whereas gaps in the leading sound release the lagging sound from ‘‘echo suppression,’’ indicating that higher-order (top–down) processes are involved in the precedence effect.

Buildup and breakdown of echo suppression for stimuli presented over headphones-the effects of interaural time and level differences.

The current study investigates buildup and breakdown of echo suppression for stimuli presented over headphones. The stimuli consisted of pairs of 120-micros clicks. The leading click (lead) and the lagging click (lag) in each pair were lateralized on opposite sides of the midline by means of interaural level differences (ILDs) of +/-10 dB or interaural time differences (ITDs) of +/-300 micros. Echo threshold was measured with an adaptive one-interval, two-alternative, forced-choice procedure with a subjective decision criterion, in which listeners had to report whether they heard a single, fused auditory event on one side of the midline, or two separate events on both sides. In the control conditions, referred to as the "single" conditions, echo threshold was measured for a single click pair, the test pair, presented in isolation. In addition to the control conditions, two kinds of test conditions were investigated, in which the test pair was preceded by 12 identical conditioning pairs: in the "same" conditions, the interaural configuration (ILDs or ITDs) of the conditioning pairs was identical to that of the test pair; in the "switch" conditions, the interaural configuration of lead and lag was reversed between the conditioning pairs and the test pair, in order to produce a switch in the lateralizations of the stimuli between the conditioning train and the test pair. No matter whether the lateralization of the clicks was produced by ILDs or by ITDs, most listeners experienced a buildup of echo suppression in the "same" conditions, manifested by a prolongation of echo threshold relative to the respective "single" conditions. However, the breakdown of echo suppression was much stronger in the ILD-switch than in the ITD-switch conditions. In five out of six listeners, the ITD switch had hardly any effect on echo threshold, although the ITDs (+/-300 micros) produced roughly the same degree of lateral displacement as the ILDs (+/-10 dB). These results suggest that the dynamic processes in echo suppression operate differentially in pathways responsible for the processing of interaural time and level differences.

Auditory gating processes and binaural inhibition in the inferior colliculus

Physiological/behavioral/perceptual responses to an auditory stimulus can be inhibited by another leading auditory stimulus at certain stimulus intervals, and have been considered useful models of auditory gating processes. Two typical examples of auditory gating are prepulse inhibition of the startle reflex and the precedence effect (echo suppression). This review summarizes studies of these two auditory gating processes with regard to their biological significance, cognitive modulation, binaural properties, and underlying neural mechanisms. Both prepulse inhibition and the precedence effect have gating functions of reducing the disruptive influence of the lagging sound, but prepulse inhibition has a much longer temporal window than the precedence effect. Attentional processes can modulate prepulse inhibition, and the listener’s previous experience can modulate the precedence effect. Compared to monaural hearing, binaural hearing reduces prepulse inhibition but enhances the precedence effect. The inferior colliculus, the major structure of the auditory midbrain, plays an important role in mediating these two auditory gating processes, and inhibitory neural transmissions within the inferior colliculus may account for binaural inhibition observed in prepulse inhibition and lag suppression recorded in the inferior colliculus. The neural mechanisms underlying binaural inhibition in the inferior colliculus are also discussed.

Functional role of the human inferior colliculus in binaural hearing

Psychophysical experiments were carried out in a rare case involving a 48 year old man (RJC) with a small traumatic hemorrhage of the right dorsal midbrain, including the inferior colliculus (IC). RJC had normal audiograms bilaterally, but there was a marked decrease in wave V amplitude on click-evoked brainstem auditory evoked potentials following left ear stimulation. RJC demonstrated a deficit in sound localization identification when the loudspeakers lay within the auditory hemifield contralateral to his IC lesion. Errors showed a consistent bias towards the hemifield ipsilateral to the lesion. Echo suppression was abnormally weak compared with that seen in control subjects, but only for sources contralateral to the lesion. Finally, speech intelligibility tests showed normal ability to benefit from spatial separation of target and competing speech sources. These results suggest that: (1) localizing sounds within a given hemifield relies on the integrity of the contralateral IC, (2) unilateral IC lesions give the illusion that sound sources in the ‘bad’ hemifield are displaced towards the ‘good’ hemifield, (3) the IC mediates aspects of echo suppression, and (4) lesion in the IC does not impede spatial release from masking in speech intelligibility, possibly due to that ability being more heavily mediated by cortical regions.

What the precedence effect tells us about room acoustics.

Sound produced in a room is typically followed by numerous reflected sounds from nearby surfaces; yet we perceive a single sound source. This perceptual phenomenon, known as the precedence effect, has long been assumed to involve echo suppression, but the nature of this suppression remains unclear. In two experiments, we investigated whether information about the lagging sound's location was perceived. Our hypothesis was that such information is critical because reflected sound can function to inform the listener about objects and structures in the room. Listeners reported hearing more echoes under a stimulus situation that simulated sudden, unexpected changes in the location of the lagging sounds, as compared with stable stimulus conditions. Placement of lagging sounds' locations proved to be critical in that a sudden shift per se did not disrupt the fusion aspect of the precedence effect; the new location had to occupy a site that specified a new reflecting surface. Perception of echoes appears to be modulated by room acoustic information contained in reflected sound and listeners' expectations about this.

Full-duplex speakerphone circuit including a supplementary echo suppressor

A supplemental echo suppressor controls the attenuators in a transmit path and receive path to selectively set the attenuations based on the signal received from the far-end and the signal transmitted from the near-end. The attenuation levels are set to allow full-duplex communication. If both ends supply speech signals simultaneously, the attenuation is reduced, increasing the clarity of signals at both ends. Attenuation levels are also set when both ends are idle. If, however, the far-end user is talking and the near-end user is silent, and the echo at the near-end is low, then the attenuation is set high. Signal parameters are generated and monitored to control echo suppression. Suppression attenuation decisions are determined on the basis of normalized power estimates that are normalized to a background power estimate that is indicative of background noise. The power and noise parameters are used to determine engagement and disengagement of the suppressors. A slow noise signal is used as a control parameter for controlling suppression. The slow noise is a secondary, lower variance background power estimate that is derived from a background power estimate. The background power estimate, in turn, is derived from a power estimate of an input signal. An Echo Return-Loss Enhancement (ERLE) expectation value is used as a control parameter to discriminate between speech signals and echo signals which, in turn, is used to make suppression gain decisions.

Echo suppression in the human cortex is affected by the spatial and temporal proximity of the primary sound and echo.

Echo suppression in the human auditory cortex was studied with auditory middle latency evoked potentials (AMEP) using virtual reality acoustic stimuli, including distance and elevation cues, presented by earphones. The purpose of the study was to assess the effect of proximity of the source sound and echo on the degree of echo suppression. Sixteen subjects were presented with source-echo pairs in which the preceding source sound was always at the vertex, and the echo varied among ten positions on the coronal plane. Positions varied in elevation, distance and time lag between source and echo. The psychoacoustic location judgment of the fused source-echo pair was closer to the source sound (more echo suppression) the nearer the echo drew to the source in its elevation and time. The equivalent dipole magnitudes of the cortical components of AMEP were significantly reduced (more suppression) with shorter echo lags and when echo elevation was similar to that of the source sound. The distances used in this study did not significantly affect echo suppression. These results indicate that echo suppression in the auditory cortex is more pronounced the closer are the primary sound and echo in locational attributes and timing. As source sound and echo draw apart, echo suppression in the cortex decreases and the perceived localization of the fused source-echo is more biased toward the echo.