Auditory Critical Bands Research Articles

Abnormal Acoustic Event Detection Based on Orthogonal Matching Pursuit in Security Surveillance System

This paper presents a feature extraction approach for surveillance system aimed at achieving the automatic detection and recognition of public security events. The proposed approach first generates a Gabor dictionary based on the human auditory critical frequency bands, and then uses the orthogonal matching pursuit (OMP) algorithm to sparse abnormal audio signal. We select the optimal several important atoms from the Gabor dictionary and extract the scale, frequency, and translation parameters of the atoms to form the OMP feature. The performance of OMP feature is compared with traditional acoustic features and their joint features, using support vector machine (SVM) and random forest (RF) classifiers. Experiments have been performed to evaluate the effectiveness of the OMP feature for supplementing traditional acoustic features. The results show the superior performance classifier for abnormal acoustic event detection (AAED) is RF. Furthermore, the introduction of the combined features addresses the problems of low recognition accuracy and poor robustness for the surveillance system in practical applications.

Sliding-band dynamic range compression for use in hearing aids

Sensorineural hearing loss is associated with elevated hearing thresholds, reduced dynamic range, and loudness recruitment. Signal processing with dynamic range compression is used in hearing aids for presenting all the sounds comfortably within the limited dynamic range of the listener. A dynamic range compression technique, named as ‘sliding-band compression’, is proposed to overcome the shortcomings of the commonly used single-band and multiband compression techniques. The gain at each frequency index is calculated as a function of the short-time power in an auditory critical band centered at it. The technique avoids the attenuation of high-frequency components due to the presence of strong low-frequency components, which may occur in single-band compression. Further, it avoids distortions in the shape of spectral resonances and discontinuities during the resonance transitions, which may occur in multiband compression. The proposed technique was implemented for offline processing and compared with single-band compression and multiband compression using objective measures and different test inputs. For two-tone inputs, the single-band compression showed attenuation of the high-frequency tone with an increase in the level of the low-frequency tone. For swept-frequency single-tone input, the multiband compression showed errors in the tone amplitude of 1.5–2.5 dB for compression ratio of 2–10. Sliding-band compression did not show either type of error. The technique has been also implemented using a fixed-point processor for real-time processing with audio latency acceptable for face-to-face communication.

REPRESENTATIONS OF DECAY IN THE WORKS OF CAT HOPE

AbstractThis article considers the ‘representation of decay’ in selected concert works by the Australian composer Cat Hope. It draws on a mixed-method research methodology, comparing the conceptual aspects of Hope's oeuvre with analyses of studio and live recordings of Hope's work and discussing how such ideas of ‘decay’ may play out in the sonic world. Two forms of spectral analysis are employed: firstly the analysis of spectral parameters roughness, noisiness, brightness, pitch, and centroid, and secondly a visualisation of the music as a spectrogram. The data for the spectral analyses are derived from Alexander Harker's spectral descriptor tools for MaxMSP which record a value for each parameter every 25 milliseconds. At times, values are normalised within a range of 0 and 1, as representative of how listeners experience parametrical changes (i.e. dynamics, in relative terms rather than absolutes in relation to other sounds in the work). Importantly, perception of noisiness is more acute at frequencies in which the auditory critical bands are wider, below 250 Hz (roughly below middle C), precisely the upper range specified by Hope to define instruments suitable for the Australian Bass Orchestra.

Neurophysiological approaches to studying the functional role of auditory critical bands

Neurophysiological data on the evaluation of neural correlates of the involvement of auditory critical bands in sound localization and recognition are presented. Single unit activity was recorded extracellularly in the central nucleus of the inferior colliculus in house mouse females. In the first part, the neuronal sensitivity to acoustic signals simulating spectral localization cues (notch noise with different bandwidths) was studied. Evaluation of neuronal response variability depending on notch bandwidths revealed that the 1/3 octave notch bandwidth was optimal for the analysis of the localization-significant spectral information. This notch bandwidth corresponded to the average critical bandwidth of the auditory midbrain neurons determined previously by the narrow-band masking method. In the second part, the selectivity of neuronal responses to a discomfort-signaling mouse pup wriggling call, its models and frequency components, was evaluated. The data obtained demonstrated that the wriggling call models with frequency components located, like in a natural call, in three non-overlapping critical bands evoked stronger neuronal responses than the other models. These findings support the suggestion of the universal role the critical-band mechanism plays in processing different parameters of acoustic stimuli.

Automatic Speaker Recognition Dependency on Both the Shape of Auditory Critical Bands and Speaker Discriminative MFCCs

Automatic Speaker Recognition Dependency on Both the Shape of Auditory Critical Bands and Speaker Discriminative MFCCs

Disordered voice measurement and auditory analysis

Although voice disorder is ordinarily first detected by listening, hearing is little used in voice measurement. Auditory critical band approaches to the quantitative analysis of dysphonia are compared with the results of applying cycle-by-cycle time based methods and the results from a listening test. The comparisons show that quite large rough/smooth differences, that are readily perceptible, are not as robustly measurable using either peripheral human hearing based GammaTone spectrograms, or a cepstral prominence algorithm, as they may be when using cycle-by-cycle based computations that are linked to temporal criteria. The implications of these tentative observations are discussed for the development of clinically relevant analyses of pathological voice signals with special reference to the analytic advantages of employing appropriate auditory criteria.

Perceptual interactions in the loudness of combined auditory and vibrotactile stimuli

The loudness of auditory (A), tactile (T), and auditory-tactile (A+T) stimuli was measured at supra-threshold levels. Auditory stimuli were pure tones presented binaurally through headphones; tactile stimuli were sinusoids delivered through a single-channel vibrator to the left middle fingertip. All stimuli were presented together with a broadband auditory noise. The A and T stimuli were presented at levels that were matched in loudness to that of the 200-Hz auditory tone at 25 dB sensation level. The 200-Hz auditory tone was then matched in loudness to various combinations of auditory and tactile stimuli (A+T), and purely auditory stimuli (A+A). The results indicate that the matched intensity of the 200-Hz auditory tone is less when the A+T and A+A stimuli are close together in frequency than when they are separated by an octave or more. This suggests that A+T integration may operate in a manner similar to that found in auditory critical band studies, further supporting a strong frequency relationship between the auditory and somatosensory systems.

Assessment of a computational model for auditory perception for rotary‐wing aircraft using human responses.

Traditional auditory perceptual models for detection of complex signals against complex ambient soundscapes are based on the human audibility threshold imposed upon computed representations of auditory critical band filters. Such models attempt to locate a positive signal‐to‐noise ratio in any critical band and then apply classic signal detection theory to derive detectability measures (d′) and probability of detection values for the event. One limitation to these models is the limited experimental validation against human sound jury performance, especially using very low‐frequency target signals such as helicopters. This study compares computational auditory detection model predictions against a corresponding large sample of human sound jury data points obtained in the laboratory. Helicopter and ambient soundscape signals were obtained from high‐sensitivity recordings in the field. Playback in the laboratory was achieved under high‐fidelity headphones calibrated to accommodate helicopter primary rotor frequencies with minimal distortion above human sensation level. All listeners completed at least 12 000 trials detecting helicopters against rural and urban soundscapes to represent the spectrum of potential environments involved in a real world scenario. Analysis compares the human sound jury performance against a contemporary computational auditory detection model, called “AUDIB,” developed by the U.S. Army and NASA.

Dichotic presentation to overcome the effect of increased spectral masking and frequency dependent hearing threshold shifts in persons with bilateral sensorineural impairment

A binaural dichotic presentation scheme for reducing the effect of increased spectral masking in persons with bilateral sensorineural loss, using spectral splitting with complementary comb filters based on auditory critical bands, has been earlier reported [Cheeran et al., J. Acoust. Soc. Am. 110, 2705 (2001)]. The 256-coefficient linear phase FIR filters designed using frequency sampling technique had transition crossovers adjusted within −6 to −4 dB for perceptual balance, and had 78–117 Hz transition, 1 dB passband ripple, and 30 dB stopband attenuation. We evaluated the scheme by conducting listening tests on 5 normal hearing subjects with simulated loss, using a closed set identification of 12 vowel-consonant-vowel syllables. Based on significant improvement, further tests were conducted on 5 hearing-impaired persons with moderate bilateral sensorineural loss. Significant improvement in response time, recognition scores, and transmission of consonantal features, particularly place and duration, was obtained, indicating reduction in the effect of spectral masking. In order to partly compensate for frequency dependent hearing threshold shifts, a pair of filters with frequency response adjusted within a 6 dB range, based on the audiogram for the corresponding ear, was cascaded with the comb filters. These filters resulted in additional improvement, particularly for persons with relatively uniform loss.

Optimal sweep cycle for time-varying comb filters for binaural dichotic presentation to improve speech perception in sensorineural hearing impairment

In a previous investigation [P. C. Pandey et al., J. Acoust. Soc. Am. 110, 2705 (2001)], a scheme using binaural dichotic presentation was devised for simultaneously reducing the effect of increased temporal and spectral masking in bilateral sensorineural hearing impairment. Speech was processed by a pair of time-varying comb filters with passbands corresponding to cyclically swept auditory critical bands, with the objective that spectral components in neighboring critical bands do not mask each other and sweeping of filter passbands provides relaxation time to the sensory cells on the basilar membrane. Presently investigation is carried out to find the optimal value of the sweep cycle. Comb filters used were 256-coefficient linear phase filters, with transition crossovers adjusted for low perceived spectral distortion, 1 dB passband ripple, 30 dB stopband attenuation, and 78–117 Hz transition width. Acoustic stimuli consisted of swept sine wave and running speech from a male and a female speaker. Bilateral loss was simulated by adding broadband noise with constant short-time SNR. Listening tests with stimuli processed using sweep cycles of 10, 20, 40, 50, 60, 80, 100 ms indicated highest perceptual quality ranking for sweep cycle in the 40–60 ms range, with a peak at 50 ms.

Comb filters for binaural dichotic presentation to improve speech perception by persons with bilateral sensorineural hearing impairment

Intelligibility of speech for persons with sensorineural hearing impairment gets reduced due to increased spectral masking. For persons with bilateral loss, improved speech perception has been reported with binaural dichotic presentation using comb filters having complementary magnitude responses, based on auditory critical bands. In this scheme, spectral components in adjacent critical bands do not mask each other as they get presented to different ears. The objective here was to design a set of comb filters with transition crossovers adjusted for a low perceived spectral distortion. For this, a slowly swept tone presented through the two comb filters should not have a change in its perceived loudness at crossover frequencies. Experiments showed that the change in loudness was negligible with crossover response lying between −4 dB and −6 dB with respect to passband response. With this constraint for response at the crossover frequencies, comb filters with responses corresponding to 18 critical bands were designed as 256-coefficient linear phase FIR filters, using frequency sampling technique. The filters have pass band ripple of 1 dB and stop band attenuation of 30 dB. These filters have resulted in improvement in speech quality and speech perception.

Binaural dichotic presentation to reduce the effect of increased temporal and spectral masking in sensorineural hearing impairment

Sensorineural hearing loss is characterized by increased temporal and spectral masking, resulting in degraded speech perception. Earlier investigations have shown that binaural dichotic presentation using comb filters with complementary magnitude responses improved speech perception, particularly the place feature. Further, inter-aural switching with periodic trapezoidal fading functions improved the perception of duration feature. For simultaneously reducing the effect of the two types of masking, a scheme has been devised with a pair of time-varying comb filters, with bands corresponding to auditory critical bands. Thus the spectral components in neighboring critical bands do not mask each other, and the sensory cells on the basilar membrane get a relaxation time due to sweeping of the filter passbands. The scheme was implemented using linear phase 256-coefficient FIR filters. Magnitude responses were cyclically swept with 20-ms period, in steps by selecting the filter coefficients for each step from a pre-calculated set. Listening tests involved closed set identification of 12 vowel–consonant–vowel syllables. The processing scheme resulted in the improvement of recognition scores, response time, and transmission of features, particularly place and duration, indicating reduction in the effect of spectral and temporal masking.

Perceptual speech coding and enhancement using frame-synchronized fast wavelet packet transform algorithms

This paper presents new wideband speech coding and integrated speech coding-enhancement systems based on frame-synchronized fast wavelet packet transform algorithms. It also formulates temporal and spectral psychoacoustic models of masking adapted to wavelet packet analysis. The algorithm of the proposed FFT-like overlapped block orthogonal wavelet packet transform permits us to efficiently approximate the auditory critical band decomposition in the time and frequency domains. This allows us to make use of the temporal and spectral masking properties of the human auditory system to decrease the average bit rate of the encoder while perceptually hiding the quantization error. The same wavelet packet representation is used to merge speech enhancement and coding in the context of auditory modeling. The advantage of the method presented in this paper over previous approaches is that perceptual enhancement and coding, which is usually implemented as a cascade of two separate systems, are combined. This leads to a decreased computational load. Experiments show that the proposed wideband coding procedure by itself can achieve transparent coding of speech signals sampled at 16 kHz at an average bit rate of 39.4 kbit/s. The combined speech coding-enhancement procedure achieves higher bit rate values that depend on the residual noise characteristics at the output of the enhancement process.

Perceptual coding of wideband audio signals: Basic principles and current applications

The need for a significant reduction in bit rate for wideband digital audio signal transmission and storage has led in the last few years to the development of psychoacoustics-based data compression techniques. In this approach, the audio frequency range is subdivided into sub-bands which are approximations of human auditory critical bands. The frequency representation of audio signals is accomplished using a filter-bank implemented as a time-to-frequency transform or as a sub-band filter. Frequency-domain masking properties of the human auditory system are exploited in order to maximize perceived fidelity of a steady-state signal at a given bit rate. In a similar way, temporal masking effects are very important in the bit-rate reduction process of transient signals. In this talk, the basic principles of perceptual audio coding will be reviewed. An adaptive transform coding technique will be presented in which both frequency-domain and temporal masking properties of human hearing are exploited during bit-rate reduction. Dolby’s current and future applications of this coding technology will be discussed.

Explaining the dispersion of the single-vowel occurrences in an F1/F2 space.

The variability of the Finnish vowel occurrences on the F1/F2 chart was studied. Repetitions (n = 8 x 10 + 10 = 160) of two syllable types, /hVh/ and /tVt/, produced by a male speaker were analysed. The F1/F2 variability in both contexts was estimated. In the /hVh/ context, the phoneme targets were considerably more peripheral than in the /tVt/ contexts. In the /tVt/ context, the phoneme placements were nearer the upper left corner of the F1/F2 space in /u/, /o/, /a/, and /ae/. That is understandable on the basis of the alveolar tongue blade position: the tongue holds its position nearer the alveolar area during its movement from [t] to [V] and back to [t] compared to the [hVh] articulation. In the latter articulation, the tongue is free to move and the vowel reaches its ideal position. This interpretation was corroborated by means of the test items with an open syllable (/tVV/). In order to explain the variability of the repetitions in one context, the notion of auditory critical band window (CBW) was used. A CBW covers an area on the psychoacoustical F1/F2 plot that comprises 1 bark on the F1 and F2 scales. Its modification CBW-F1 implies that a critical band is presented as a circle comprising the diameter of 1 bark according to the F1 scale. Practically all single F1/F2 points occur within the CBW-F1 circle. In most cases, the variability is smaller than one CBW-F1 circle, but we can consider that a greater amount of repetitions would yield a greater dispersion.(ABSTRACT TRUNCATED AT 250 WORDS)

Classification of background noises for hearing-aid applications.

A background-noise classification procedure is being developed for hearing-aid applications, wherein the hearing-aid response would be adjusted in response to changes in the noise environment. The classification procedure is based on measuring four signal features giving the fluctuations of the signal envelope and the mean frequency and low- and high-frequency slopes of the average spectrum. A more complicated procedure, based on determining the envelope modulation spectra in auditory critical bands, was also investigated and was found to offer no advantages over the simpler procedure. The accuracy of the classification procedure was determined for eleven everyday background noises under optimal conditions where the training and test noise sequences were different portions of the same short noise recording. A cluster analysis was used to determine the similarities among the feature vectors for the noises, and when the noises are grouped into a reduced number of clusters the noise-classification accuracy using the four features exceeds 90%.

Efficient representation of LPC excitation using nonuniform frequency-domain sampling.

An efficient representation of the LPC excitation is essential in predictive coding systems for synthesizing high-quality speech at low bit rates. In this paper, a method is presented that takes advantage of the nonuniform spacing of auditory critical bands to achieve an efficient frequency-domain representation of LPC excitation. A segment of LPC excitation with a duration of N samples, represented as a Fourier series, requires N/2 sinusoidal components uniformly spaced along the frequency axis for exact reproduction of the excitation. Thus, for speech bandlimited to 4 kHz, a 10-ms segment requires 40 frequency components for exact reproduction. It was found that, by using uniform frequency spacing below 1 kHz and logarithmic spacing above 1 kHz, the number of sinusoidal components can be reduced to 15 without introducing any audible distortion in the synthetic speech signal. Subjective tests were conducted to determine the effective signal-to-noise ratio of synthetic speech for different numbers of sinusoidal components. These results and a tape demonstration of synthetic speech will be presented at the meeting.

Estimation of the acoustical transfer function from vowel confusions.

An attempt is made to estimate a spectral alteration in the acoustical transfer function (including the transmission characteristic of the peripheral auditory system) from vowel confusions. To study the relation between a predetermined spectral alteration and the resulting vowel confusions in a closed nonsense-syllable test, different conditions of filtered speech stimuli were employed with and without interfering speech-spectrum-shaped noise. The number of available perceptional cues and the variability of the material was restricted by synthesizing speech with a LAR vocoder. The vowel confusions between 14 syllables were analyzed with the transinformation analysis for the different filtered conditions. For this purpose, a set of 15 acoustical features was constructed. Each feature represents the average log power within one out of 15 auditory critical bands. Using these features, the ‘‘transinformation spectrum’’ is computed, i.e., the transmitted speech information as a function of frequency.This quantity might be used to predict the acoustical transfer function. Based on the empirical results, the correspondence between alterations in the acoustical transfer function and changes in the transinformation spectrum will be discussed.

Comparison of auditory critical bands measured by loudness summation and acoustic reflexes

The critical bandwidth in four normal hearing listeners was examined using two different psychophysical methods. Loudness balances and acoustic reflex measures were made to a two-tone stimulus when the components comprising the complex were adjusted to an “equal intensity” level or an “equal reflex” level. Measurements following these adjustments facilitated the comparison of two-tone loudness summation with two-tone acoustic reflex threshold as a function of sound pressure level. Measurements were made centered around 1000 Hz for various bandwidth spacings of the two-tone complex ranging from 100 to 3000 Hz. The results indicate a difference in critical bandwidths which depends upon the method of adjustment of the two-tone complex (equal reflex vs. equal intensity) for both the loudness balance and the acoustic reflex tasks. By using similar durations, similar sound pressure levels, and similar adjustments of the two-tone complex the comparison of the two-tone acoustic reflex to the loudness of the eliciting signal was facilitated. For this procedure the acoustic reflex appeared to be loudness determined, at least at reflex threshold levels.