Spectral Contrast Enhancement Research Articles

Effect of Spectral Contrast Enhancement on Speech-on-Speech Intelligibility and Voice Cue Sensitivity in Cochlear Implant Users.

Speech intelligibility in the presence of a competing talker (speech-on-speech; SoS) presents more difficulties for cochlear implant (CI) users compared with normal-hearing listeners. A recent study implied that these difficulties may be related to CI users' low sensitivity to two fundamental voice cues, namely, the fundamental frequency (F0) and the vocal tract length (VTL) of the speaker. Because of the limited spectral resolution in the implant, important spectral cues carrying F0 and VTL information are expected to be distorted. This study aims to address two questions: (1) whether spectral contrast enhancement (SCE), previously shown to enhance CI users' speech intelligibility in the presence of steady state background noise, could also improve CI users' SoS intelligibility, and (2) whether such improvements in SoS from SCE processing are due to enhancements in CI users' sensitivity to F0 and VTL differences between the competing talkers. The effect of SCE on SoS intelligibility and comprehension was measured in two separate tasks in a sample of 14 CI users with Cochlear devices. In the first task, the CI users were asked to repeat the sentence spoken by the target speaker in the presence of a single competing talker. The competing talker was the same target speaker whose F0 and VTL were parametrically manipulated to obtain the different experimental conditions. SoS intelligibility, in terms of the percentage of correctly repeated words from the target sentence, was assessed using the standard advanced combination encoder (ACE) strategy and SCE for each voice condition. In the second task, SoS comprehension accuracy and response times were measured using the same experimental setup as in the first task, but with a different corpus. In the final task, CI users' sensitivity to F0 and VTL differences were measured for the ACE and SCE strategies. The benefit in F0 and VTL discrimination from SCE processing was evaluated with respect to the improvement in SoS perception from SCE. While SCE demonstrated the potential of improving SoS intelligibility in CI users, this effect appeared to stem from SCE improving the overall signal to noise ratio in SoS rather than improving the sensitivity to the underlying F0 and VTL differences. A second key finding of this study was that, contrary to what has been observed in a previous study for childlike voice manipulations, F0 and VTL manipulations of a reference female speaker (target speaker) toward male-like voices provided a small but significant release from masking for the CI users tested. The present findings, together with those previously reported in the literature, indicate that SCE could serve as a possible background-noise-reduction strategy in commercial CI speech processors that could enhance speech intelligibility especially in the presence of background talkers that have longer VTLs compared with the target speaker.

Evaluation of combined dynamic compression and single channel noise reduction for hearing aid applications

Objective: Single-channel noise reduction (SCNR) and dynamic range compression (DRC) are important elements in hearing aids. Only relatively few studies have addressed interaction effects and typically used real hearing aids with limited knowledge about the integrated algorithms. Here the potential benefit of different combinations and integration of SCNR and DRC was systematically assessed. Design: Ten different systems combining SCNR and DRC were implemented, including five serial arrangements, a parallel and two multiplicative approaches. In an instrumental evaluation, signal-to-noise ratio (SNR) improvement and spectral contrast enhancement (SCE) were assessed. Quality ratings at 0 and +6 dB SNR, and speech reception thresholds (SRTs) in noise were measured using stationary and babble noise. Study sample: Thirteen young normal-hearing (NH) listeners and 12 hearing-impaired (HI) listeners participated. Results: In line with an increased segmental SNR and spectral contrast compared to a serial concatenation, the parallel approach significantly reduced the perceived noise annoyance for both subject groups. The proposed multiplicative approaches could partly counteract increased speech distortions introduced by DRC and achieved the best overall quality for the HI listeners. Conclusions: For high SNRs well above the individual SRT, the specific combination of SCNR and DRC is perceptually relevant and the integrative approaches were preferred.

Enhancement of spectral contrast and spectral changes as approaches to improving the intelligibility of speech in background sounds

Cochlear hearing loss is accompanied by reduced frequency selectivity, which contributes to problems in understanding speech in background sounds. The excitation pattern evoked by sounds like vowels is “flatter” in an impaired ear than in a normal ear. Two approaches to partially compensating for the effects of reduced frequency selectivity are described. The first uses spectral contrast enhancement on a frame-by-frame basis. Only features of the short-term spectrum that would be represented in a normal ear are enhanced. Experimental evaluations of this approach showed that it could improve the intelligibility of speech in noise by a small amount, but that this required a fine frequency analysis, in turn requiring relatively long frames and time delays that might be too long for use of the method in hearing aids. A second approach enhances spectral changes over time, based on the idea that information is speech is carried by spectral changes. This approach also led to modest improvements in the intelligibility of speech in noise, especially when the amount of enhancement was made to vary with frequency depending on the hearing loss at that frequency and when the parameters of the processing were selected for each individual using a genetic algorithm.

Spectral contrast enhancement improves speech intelligibility in noise for cochlear implants.

Spectral smearing causes, at least partially, that cochlear implant (CI) users require a higher signal-to-noise ratio to obtain the same speech intelligibility as normal hearing listeners. A spectral contrast enhancement (SCE) algorithm has been designed and evaluated as an additional feature for a standard CI strategy. The algorithm keeps the most prominent peaks within a speech signal constant while attenuating valleys in the spectrum. The goal is to partly compensate for the spectral smearing produced by the limited number of stimulation electrodes and the overlap of electrical fields produced in CIs. Twelve CI users were tested for their speech reception threshold (SRT) using the standard CI coding strategy with and without SCE. No significant differences in SRT were observed between conditions. However, an analysis of the electrical stimulation patterns shows a reduction in stimulation current when using SCE. In a second evaluation, 12 CI users were tested in a similar configuration of the SCE strategy with the stimulation being balanced between the SCE and the non-SCE variants such that the loudness perception delivered by the strategies was the same. Results show a significant improvement in SRT of 0.57 dB (p < 0.0005) for the SCE algorithm.

Optimization of a Spectral Contrast Enhancement Algorithm for Cochlear Implants Based on a Vowel Identification Model.

Speech intelligibility achieved with cochlear implants (CIs) shows large variability across different users. One reason that can explain this variability is the CI user's individual electrode nerve interface which can impact the spectral resolution they can achieve. Spectral resolution has been reported to be related to vowel and consonant recognition in CI listeners. One measure of spectral resolution is the spectral modulation threshold (SMT), which is defined as the smallest detectable spectral contrast in a stimulus. In this study we hypothesize that an algorithm that improves SMT may improve vowel identification, and consequently produce an improvement in speech understanding for CIs. With this purpose we implemented an algorithm, termed spectral contrast enhancement (SCE) that emphasizes peaks with respect to valleys in the audio spectrum. This algorithm can be configured with a single parameter: the amount of spectral contrast enhancement entitled "SCE factor". We would like to investigate whether the "SCE factor" can be individualized to each CI user. With this purpose we used a vowel identification model to predict the performance produced by the SCE algorithm with different "SCE factors" in a vowel identification task.In five CI users the new algorithm has been evaluated using a SMT task and a vowel identification task. The tasks were performed for SCE factors of 0 (no enhancement), 2 and 4. In general it seems that increasing the SCE factor produces a decrease in performance in both the SMT threshold and vowel identification.

A quantitative analysis of spectral mechanisms involved in auditory detection of coloration by a single wall reflection

Coloration detection thresholds (CDTs) were measured for a single reflection as a function of spectral content and reflection delay for diotic stimulus presentation. The direct sound was a 320-ms long burst of bandpass-filtered noise with varying lower and upper cut-off frequencies. The resulting threshold data revealed that: (1) sensitivity decreases with decreasing bandwidth and increasing reflection delay and (2) high-frequency components contribute less to detection than low-frequency components. The auditory processes that may be involved in coloration detection (CD) are discussed in terms of a spectrum-based auditory model, which is conceptually similar to the pattern-transformation model of pitch ( Wightman, 1973). Hence, the model derives an auto-correlation function of the input stimulus by applying a frequency analysis to an auditory representation of the power spectrum. It was found that, to successfully describe the quantitative behavior of the CDT data, three important mechanisms need to be included: (1) auditory bandpass filters with a narrower bandwidth than classic Gammatone filters, the increase in spectral resolution was here linked to cochlear suppression, (2) a spectral contrast enhancement process that reflects neural inhibition mechanisms, and (3) integration of information across auditory frequency bands.

Spectral contrast enhancement: Algorithms and comparisons

This paper investigates spectral contrast enhancement techniques and their implementation complexity. Three algorithms are dealt with in this paper. The first is the method described by Baer, Moore and Gatehouse. Two alternative methods are also proposed and investigated in this paper from a practical application and implementation point of view. Theoretical analyses and results from laboratory, simulation and subject listening show that spectral contrast enhancement and performance improvement can be achieved by use of these three methods with the appropriate selection of their relevant parameters.

Minimum spectral contrast needed for vowel identification by normal hearing and cochlear implant listeners.

The minimum spectral contrast needed for vowel identification by normal-hearing and cochlear implant listeners was determined in this study. In experiment 1, a spectral modification algorithm was used that manipulated the channel amplitudes extracted from a 6-channel continuous interleaved sampling (CIS) processor to have a 1-10 dB spectral contrast. The spectrally modified amplitudes of eight natural vowels were presented to six Med-EI/CIS-link users for identification. Results showed that subjects required a 4-6 dB contrast to identify vowels with relatively high accuracy. A 4-6 dB contrast was needed independent of the individual subject's dynamic range (range 9-28 dB). Some cochlear implant (CI) users obtained significantly higher scores with vowels enhanced to 6 dB contrast compared to the original, unenhanced vowels, suggesting that spectral contrast enhancement can improve the vowel identification scores for some CI users. To determine whether the minimum spectral contrast needed for vowel identification was dependent on spectral resolution (number of channels available), vowels were processed in experiment 2 through n (n =4, 6, 8, 12) channels, and synthesized as a linear combination of n sine waves with amplitudes manipulated to have a 1-20 dB spectral contrast. For vowels processed through 4 channels, normal-hearing listeners needed a 6 dB contrast, for 6 and 8 channels a 4 dB contrast was needed, consistent with our findings with CI listeners, and for 12 channels a 1 dB contrast was sufficient to achieve high accuracy (>80%). The above-mentioned findings with normal-hearing listeners suggest that when the spectral resolution is poor, a larger spectral contrast is needed for vowel identification. Conversely, when the spectral resolution is fine, a small spectral contrast (1 dB) is sufficient. The high identification score (82%) achieved with 1 dB contrast was significantly higher than any of the scores reported in the literature using synthetic vowels, and this can be attributed to the fact that we used natural vowels which contained duration and spectral cues (e.g., formant movements) present in fluent speech. The outcomes of experiments 1 and 2, taken together, suggest that CI listeners need a larger spectral contrast (4-6 dB) than normal-hearing listeners to achieve high recognition accuracy, not because of the limited dynamic range, but because of the limited spectral resolution.

The effect of spectral contrast enhancement on speech perception in hearing-impaired listeners

The effect of enhancing the spectral contrast of clean and noisy speech on intelligibility and listening comfort in hearing-impaired listeners is explored. Spectral contrast refers to the difference between the amplitude of spectral peaks and valleys in a short temporal window. A method of enhancing spectral contrast similar to that suggested by Bunnell [H. T. Bunnell, J. Acoust. Soc. Am. 88, 2546–2556 (1990)] was implemented in a simulated hearing aid with fast acting multiband wide-dynamic-range compression. The aid’s frequency response was adjusted to accommodate the individual listeners’ hearing losses. The effect of several degrees of spectral contrast enhancement on phoneme identification in quiet and in the presence of speech-spectrum-shaped noise was measured. In addition, listeners’ preferences for several degrees of spectral contrast enhancement were assessed in an A/B comparison of spectrally enhanced connected speech in quiet and in the presence of cafeteria noise. The experimental results will be reported and their implications for the use of spectral contrast enhancements in hearing aids will be discussed.

Growth of spectral contrast enhancement in Schroeder-phase harmonic complexes

Discrimination of spectral differences in harmonic complexes is better for waveforms that are highly modulated and presented at high levels relative to lower-level stimuli with flatter temporal envelopes [M. R. Leek and V. Summers, J. Acoust. Soc. Am. 94, 2074–2082 (1993)]. This may reflect an enhancement of contrast between spectral peaks and valleys related to nonlinear cochlear processing. To further explore this relationship, discrimination of harmonic complexes with slightly different peak frequencies was measured at several stimulus levels. Peak-to-background differences required for discrimination were taken as a measure of the spectral contrast preserved in the auditory system. Phases were selected to generate either flat or peaked internal waveforms (negative or positive Schroeder phase), and peak frequency regions were near 2, 3, or 4 kHz. In normal-hearing listeners, the positive Schroeder waveforms produced lower spectral contrast thresholds than the negative Schroeder stimuli at all tested intensities and peak regions, with the greatest differences at moderate stimulus levels. Generally, hearing-impaired listeners demonstrated smaller threshold differences due to phase selection. These results are consistent with normal cochlear processing that is nonlinear at moderate levels, but more linear at low and high levels, and with more linear processing in impaired ears. [Work supported by NIH.]

Spectral enhancement using a two-filter model of impaired frequency selectivity

Neural tuning curves can be described as comprising two shapes: a narrow ‘‘tip’’ filter with high sensitivity, and a wider ‘‘tail’’ filter with lower sensitivity. When cochlear function is impaired, the sensitivity of the tip filter decreases. The tail filter characteristics are difficult to measure in normal-hearing subjects using psychophysical techniques such as the notched-noise method. However, with mild to severe cochlear hearing loss, the tail filter has a greater influence on the results, and becomes easier to measure. Auditory filter shapes were estimated using the notched-noise method over a wide range of center frequencies for subjects with differing degrees of hearing loss. These data were used to construct a model that predicts typical filter shapes as a function of frequency and hearing loss. The model is being used to improve the accuracy of a spectral contrast enhancement method intended to compensate for the effects of reduced frequency selectivity. The method is similar to one also being developed in this laboratory [T. Baer and B. C. J. Moore, this meeting], but uses a filter bank rather than overlap-add FFTs. Results of intelligibility tests using speech in speech-shaped noise, enhanced by this method and tested on hearing-impaired subjects, will be described. [Work supported by the MRC.]