A cocktail party model of spatial release from masking by both noise and speech interferers

A mathematical formula for estimating spatial release from masking (SRM) in a cocktail party environment would be useful as a simpler alternative to computationally intensive algorithms and may enhance understanding of underlying mechanisms. The experiment presented herein was designed to provide a strong test of a model that divides SRM into contributions of asymmetry and angular separation [Bronkhorst (2000). Acustica 86, 117-128] and to examine whether that model can be extended to include speech maskers. Across masker types the contribution to SRM of angular separation of maskers from the target was found to grow at a diminishing rate as angular separation increased within the frontal hemifield, contrary to predictions of the model. Speech maskers differed from noise maskers in the overall magnitude of SRM and in the contribution of angular separation (both greater for speech). These results were used to develop a modified model that achieved good fits to data for noise maskers (ρ=0.93) and for speech maskers (ρ=0.94) while using the same functions to describe separation and asymmetry components of SRM for both masker types. These findings suggest that this approach can be used to accurately model SRM for speech maskers in addition to primarily "energetic" noise maskers.     

Monaural masking release in random-phase and low-noise noise

Two masking-release paradigms thought to involve across-channel processing are comodulation masking release (CMR) and profile analysis. Similarities between these two paradigms were explored by comparing signal detection in maskers that varied only in degree of envelope fluctuation. The narrow-band-noise maskers were 10 Hz wide and their envelope fluctuations were manipulated using the low-noise noise algorithm of Pumplin [J. Acoust. Soc. Am. 78, 100-104 (1985)]. Masking conditions included the classic CMR conditions of an on-frequency band, multiple (five) incoherent bands, or multiple coherent bands. Detection was compared using both random-phase noise (RPN) and low-noise noise (LNN) maskers. In one set of conditions, the signal was identical to the on-frequency masker, yielding an intensity discrimination task. Conditions that included RPN maskers and tonal signals resembled the classic CMR paradigm, whereas conditions including LNN and noise signals more closely resembled the classic profile analysis paradigm. Other conditions may be considered hybrids. This combination of conditions provided a wide variety of within- and across-channel cues for detection. The results suggest that CMR and profile analysis could be based upon the same set of stimulus cues and perhaps the same perceptual processes.     

Free-field release from masking

Free-field release from masking was studied as a function of the spatial separation of a signal and masker in a two-interval, forced-choice (2IFC) adaptive paradigm. The signal was a 250-ms train of clicks (100/s) generated by filtering 50-microseconds pulses with a TDH-49 speaker (0.9 to 9.0 kHz). The masker was continuous broadband (0.7 to 11 kHz) white noise presented at a level of 44 dBA measured at the position of the subject's head. In experiment I, masked and absolute thresholds were measured for 36 signal source locations (10 degree increments) along the horizontal plane as a function of seven masking source locations (30 degree increments). In experiment II, both absolute and masked thresholds were measured for seven signal locations along three vertical planes located at azimuthal rotations of 0 degrees (median vertical plane), 45 degrees, and 90 degrees. In experiment III, monaural absolute and masked thresholds were measured for various signal-masker configurations. Masking-level differences (MLDs) were computed relative to the condition where the signal and mask were in front of the subjects after using absolute thresholds to account for differences in the signal's sound-pressure level (SPL) due to direction. Maximum MLDs were 15 dB along the horizontal plane, 8 dB along the vertical, and 9 dB under monaural conditions.     

Frequency specificity of the human auditory brainstem and middle latency responses using notched noise masking

This study investigated the frequency specificity of the auditory brainstem and middle latency responses to 80 and 90 dB ppe SPL 500-Hz and 90 dB ppe SPL 2000-Hz tonebursts. The stimuli were brief (2-1-2 cycle) linear-gated tonebursts. ABR/MLRs were recorded using two electrode montages: (1) Cz-nape of neck and (2) Cz-ipsilateral earlobe. Cochlear contributions to ABR wave V-Na and MLR waves Na-Pa and Pa-Nb were assessed by plotting notched noise tuning curves which showed amplitudes and latencies as a function of center frequency of the noise masker [Abdala and Folsom, J. Acoust. Soc. Am. 97, 2394 (1995); ibid. 98, 921 (1995)]. Maxima in the response amplitude profiles for the ABR and MLR to 80 dB ppe SPL tonebursts occurred within one-half octave of the nominal stimulus frequency, with minimal contributions to the responses from frequencies greater than one octave away. At 90 dB ppe SPL, contributions came from a slightly broader frequency region for both stimulus frequencies. Thus, the ABR/MLR to 80 dB ppe SPL tonebursts shows good frequency specificity which decreases at 90 dB ppe SPL. No significant differences exist in frequency specificity of: (1) ABR wave V-Na versus MLR waves Na-Pa and Pa-Nb at either stimulus frequency or intensity; and (2) ABR/MLRs recorded using the two electrode montages.     

Sentence recognition in noise promoting or suppressing masking release by normal-hearing and cochlear-implant listeners

Normal-hearing (NH) listeners maintain robust speech understanding in modulated noise by "glimpsing" portions of speech from a partially masked waveform--a phenomenon known as masking release (MR). Cochlear implant (CI) users, however, generally lack such resiliency. In previous studies, temporal masking of speech by noise occurred randomly, obscuring to what degree MR is attributable to the temporal overlap of speech and masker. In the present study, masker conditions were constructed to either promote (+MR) or suppress (-MR) masking release by controlling the degree of temporal overlap. Sentence recognition was measured in 14 CI subjects and 22 young-adult NH subjects. Normal-hearing subjects showed large amounts of masking release in the +MR condition and a marked difference between +MR and -MR conditions. In contrast, CI subjects demonstrated less effect of MR overall, and some displayed modulation interference as reflected by poorer performance in modulated maskers. These results suggest that the poor performance of typical CI users in noise might be accounted for by factors that extend beyond peripheral masking, such as reduced segmental boundaries between syllables or words. Encouragingly, the best CI users tested here could take advantage of masker fluctuations to better segregate the speech from the background.     

Neural correlates of psychophysical release from masking

Responses of chinchilla auditory-nerve fibers were measured for stimulus conditions analogous to those in which psychophysical release from masking has been observed in humans. The maskers were two equal power, narrow-band noise stimuli with different amplitude envelopes. The neurons in the sample fell into three groups that resolved the maskers' envelopes with varying degrees of accuracy. The boundaries of these groups were not sharply delineated by characteristic frequency (CF) but were dependent on the relationship between the masker level and the neurons' thresholds at the masker frequency. For the neurons that best preserved the maskers' envelope fluctuations, a neural release from masking was observed; rate-based neural masked thresholds were higher for the masker with the least fluctuating envelope. The results suggest that neural and psychophysical release from masking arises because the probe evokes larger rate changes, relative to the background response to the masker, during periods of low masker energy. Between two otherwise equivalent maskers, the one with the periods of lowest energy will produce the lower masked thresholds because rate changes are larger and more detectable.     

Stimulus factors influencing spatial release from speech-on-speech masking

This study examined spatial release from masking (SRM) when a target talker was masked by competing talkers or by other types of sounds. The focus was on the role of interaural time differences (ITDs) and time-varying interaural level differences (ILDs) under conditions varying in the strength of informational masking (IM). In the first experiment, a target talker was masked by two other talkers that were either colocated with the target or were symmetrically spatially separated from the target with the stimuli presented through loudspeakers. The sounds were filtered into different frequency regions to restrict the available interaural cues. The largest SRM occurred for the broadband condition followed by a low-pass condition. However, even the highest frequency bandpass-filtered condition (3-6 kHz) yielded a significant SRM. In the second experiment the stimuli were presented via earphones. The listeners identified the speech of a target talker masked by one or two other talkers or noises when the maskers were colocated with the target or were perceptually separated by ITDs. The results revealed a complex pattern of masking in which the factors affecting performance in colocated and spatially separated conditions are to a large degree independent.     

Simultaneous masking by gated and continuous sinusoidal maskers

Simultaneous masking of a 20-ms, 1-kHz signal was investigated using 50-ms gated and continuous sinusoidal maskers with frequencies below, at, and above 1 kHz. Gated maskers can produce considerably (5-20 dB) more masking than continuous maskers, and this difference does not appear to result from the spread of energy produced by gating either the masker or the signal. For masker frequencies below the signal frequency, this difference in masking is primarily due to the detection of the cubic difference tone in the continuous condition. For masker frequencies at and above the signal frequency, the difference appears to be an important property of masking. Implications of this frequency-dependent effect for measures of frequency selectivity are discussed.     

Factors affecting masking release for speech in modulated noise for normal-hearing and hearing-impaired listeners

The Speech Reception Threshold for sentences in stationary noise and in several amplitude-modulated noises was measured for 8 normal-hearing listeners, 29 sensorineural hearing-impaired listeners, and 16 normal-hearing listeners with simulated hearing loss. This approach makes it possible to determine whether the reduced benefit from masker modulations, as often observed for hearing-impaired listeners, is due to a loss of signal audibility, or due to suprathreshold deficits, such as reduced spectral and temporal resolution, which were measured in four separate psychophysical tasks. Results show that the reduced masking release can only partly be accounted for by reduced audibility, and that, when considering suprathreshold deficits, the normal effects associated with a raised presentation level should be taken into account. In this perspective, reduced spectral resolution does not appear to qualify as an actual suprathreshold deficit, while reduced temporal resolution does. Temporal resolution and age are shown to be the main factors governing masking release for speech in modulated noise, accounting for more than half of the intersubject variance. Their influence appears to be related to the processing of mainly the higher stimulus frequencies. Results based on calculations of the Speech Intelligibility Index in modulated noise confirm these conclusions.     

Spatial release from masking of aerial tones in pinnipeds

In most masking experiments, target signals and sound intended to mask are located in the same position. Spatial release from masking (SRM) occurs when signals and maskers are spatially separated, resulting in detection improvement relative to when they are spatially co-located. In this study, SRM was investigated in a harbor seal, who naturally lacks pinnae, and California sea lion, who possesses reduced pinnae. Subjects had to detect aerial tones at 1, 8, and 16 kHz in the presence of octave bands of white noise centered at the tone frequency. While the masker occurred in front of the subject (0 degree), the tone occurred at 0, 45, or 90 degrees in the horizontal plane. Unmasked thresholds were also measured at these angles to determine sensitivity differences based on source azimuth. Compared to when signal and masker where co-located, masked thresholds were lower by as much as 19 and 12 dB in the harbor seal and sea lion, respectively, when signal and masker were separated. Masked threshold differences of the harbor seal were larger than those previously measured under water. Performance was consistent with some measurements collected on terrestrial animals but differences between subjects at the highest frequency likely reflect variations in pinna anatomy.     

Multiple bursts, multiple looks, and stream coherence in the release from informational masking

In the simultaneous multitone masking paradigm introduced by Neff and Green [Percept. Psychophys. 41, 409-415 (1987)] the masker typically is a small number of tones having frequencies and levels that are randomly drawn on every presentation. Large amounts of masking for a pure-tone signal often occur that are thought to reflect central, rather than peripheral, limitations on processing. Previous work from this laboratory has indicated that playing a rapid succession of randomly drawn multitone maskers in each observation interval dramatically reduces the amount of masking that is observed relative to a single burst (SB). In this multiple-bursts-different (MBD) procedure, the signal tone is the only constant frequency component during the sequence of bursts and tends to perceptually segregate from the masker. In this study, the number of masker bursts and the interburst interval (IBI) were varied. The goals were to determine how the release from masking relative to the SB condition depends on the number of bursts and to examine whether increasing the IBI would cause each burst to be processed independently. If the latter were true, it might disrupt the perception of signal stream coherence, thereby diminishing the MBD advantage. However, multiple independent looks could also lead to an improvement in performance. For those subjects showing large amounts of informational masking in the SB condition, substantial reduction in masked thresholds occurred as the number of masker bursts increased, while masking increased as IBI lengthened. The results were not consistent with a simple version of a multiple-look model in which the information from each burst was combined optimally, but instead appear to be attributable to mechanisms involved in the perceptual organization of sounds.     

Spatial release from speech-on-speech masking in the median sagittal plane

Several studies have described a release from speech-on-speech masking associated with separation of target and masker sources in the median sagittal plane. Some have excluded the possibility that small differences between target and masker interaural time disparities can fully account for this release. This study explored the mechanisms underlying the spatial release from speech-on-speech masking that can be obtained in the absence of such differences. In one condition, interaural time disparities were removed from the nominal median-sagittal-plane, head-related impulse responses used to generate the virtual auditory space within which competing sentences were presented. In other conditions, interaural level and spectral disparities also were manipulated by presenting competing sentences monaurally or diotically after convolution with one ear's head-related impulse responses. It was found that substantial spatial release from masking can be obtained in the absence of any interaural disparities and that such disparities probably make a relatively minor contribution to spatial release from speech-on-speech masking in the median sagittal plane. It is argued that this release from masking is driven primarily by a reduction in informational masking that occurs when monaural information at one, or both, of the listener's ears facilitates differentiation of competing sentences that emanate from spatially separated sources.     

Speech intelligibility and spatial release from masking in young children

Children between the ages of 4 and 7 and adults were tested in free field on speech intelligibility using a four-alternative forced choice paradigm with spondees. Target speech was presented from front (0 degrees); speech or modulated speech-shaped-noise competitors were either in front or on the right (90 degrees). Speech reception thresholds were measured adaptively using a three-down/one-up algorithm. The primary difference between children and adults was seen in elevated thresholds in children in quiet and in all masked conditions. For both age groups, masking was greater with the speech-noise versus speech competitor and with two versus one competitor(s). Masking was also greater when the competitors were located in front compared with the right. The amount of masking did not differ across the two age groups. Spatial release from masking was similar in the two age groups, except for in the one-speech condition, when it was greater in children than adults. These findings suggest that, similar to adults, young children are able to utilize spatial and/or head shadow cues to segregate sounds in noisy environments. The potential utility of the measures used here for studying hearing-impaired children is also discussed.     

A novel feedback active noise control for broadband chaotic noise and random noise

The feedback active noise control (ANC) can be seen as a predictor, the conventional method based on filtered-x least mean square (FXLMS) algorithm can only be useful for linear and tonal noise, but for nonlinear and broadband noise, it is useless. The feedback ANC using functional link artificial neural networks (FLANN) based on filtered-s least mean square (FSLMS) algorithm can reduce some nonlinear noise such as chaotic noise, but the noise cancellation performance is not very well, at the same time, it is not useful to random noise. To solve the problem above, a new feedback ANC using wavelet packet FXLMS (WPFXLMS) algorithm is proposed in this paper. By decomposing the broadband noise into several band-limited parts which are predictable and each part is controlled independently, the proposed algorithm can not only suppress the chaotic noise, but also mitigate the random noise. Compared with FXLMS and FSLMS algorithms, proposed WPFXLMS algorithm also holds the best performance on noise cancellation. Numerous simulations are conducted to demonstrate the effectiveness of the proposed WPFXLMS algorithm.     

Pitch shift of pure and complex tones induced by masking noise

Psychoacoustic experiments were performed to measure the pitch-shift effects of pure and complex tones resulting from the addition of a masking noise to the tonal stimuli. Harmonic residue tones with either two or three harmonics and a fundamental frequency of 200 Hz were chosen as test tones. The pitch shifts of virtual and spectral pitches of the residue tones were measured as a function of the intensity of a low-pass noise with 600-Hz cutoff frequency. The SPL of this noise varied between 30 and 70 dB. In another experiment, the pitch shifts of single pure tones corresponding to the frequencies and SPLs of the harmonics of the residue tones were measured using the same masking noise. The results from five subjects for the harmonic residue tones show only a weak dependence of pitch shift on masking noise intensity. This dependence exists for both spectral and virtual pitches. In the case of single pure tones, pitch shift depends more distinctly on noise intensity. Pitch shifts of up to 5% were found in the range of noise intensity investigated. The magnitude of pitch shift shows pronounced interindividual differences, but the direction of the shift effect is always the same. In all cases pitch increases with higher masking noise levels.