Spectral Modulation Detection Research Articles

Spectral modulation detection training in older adults with hearing loss.

Individuals with hearing loss are poorer than normal‐hearing listeners at detecting patterns of peaks and valleys of sound level spread across audio frequency (spectral modulation detection)—an important skill for speech and music perception. In this preliminary report we examined whether listeners with hearing loss could improve on spectral modulation detection with practice. We trained older adults with sensorineural hearing loss (n=7) ∼1 hr/day for 7 days to distinguish a 400–3200 Hz noiseband with a flat‐spectrum from one with a 2 cyc/oct sinusoidal spectral modulation. Modulation depth was varied adaptively to determine the detection threshold. Their mean thresholds improved significantly from 18.9 to 12.8 dB. These listeners also improved at both a lower and a higher untrained spectral modulation frequency (1 and 4 cyc/oct), but showed no change on a measure of frequency selectivity. Interestingly, a separate group of younger normal‐hearing adults who received the same training (n=7) did not improve on the trained (6.9 to 6.6 dB) or untrained conditions. Normal‐hearing listeners had lower pre‐training thresholds than individuals with hearing loss, and across all listeners the magnitude of individual improvement was correlated to pre‐training threshold (r=0.92). Listener age and/or hearing loss could have contributed to these across‐population differences.

Spectral modulation detection training: Different outcomes for different trained stimuli.

The pattern of peaks and valleys of sound level spread across audio frequency (spectral modulation) is a key stimulus cue used for sound identification and localization. However, little is known about how practice can influence the ability to detect this modulation. Here we trained three separate groups of normal‐hearing adults to detect one of three different spectral modulations. Listeners practiced distinguishing a noise with a flat spectrum from one with a sinusoidal spectral modulation 1 hr/day for 7 days. We varied the modulation depth adaptively to determine the detection threshold. On average, the groups trained to detect 0.5 cyc/oct or 1 cyc/oct spectral modulation across a 200–1600 Hz carrier improved significantly, while the group trained to detect 2 cyc/oct spectral modulation across a 400–3200 Hz carrier did not. Further, it took multiple sessions for the 0.5‐cyc/oct group to reach optimal performance, but only one session for the 1‐cyc/oct group. Thus it appears that the current training had a smaller and more rapid effect as spectral modulation frequency increased from 0.5 to 2 cyc/oct (if we assume no influence of carrier). Pre‐training thresholds also decreased over this range, suggesting that there is less room to improve at higher spectral modulation frequencies.

Spectral modulation masking patterns reveal tuning to spectral envelope frequency

Auditory processing appears to include a series of domain-specific filtering operations that include tuning in the audio-frequency domain, followed by tuning in the temporal modulation domain, and perhaps tuning in the spectral modulation domain. To explore the possibility of tuning in the spectral modulation domain, a masking experiment was designed to measure masking patterns in the spectral modulation domain. Spectral modulation transfer functions (SMTFs) were measured for modulation frequencies from 0.25 to 14 cycles/octave superimposed on noise carriers either one octave (800-1600 Hz, 6400-12,800 Hz) or six octaves wide (200-12,800 Hz). The resulting SMTFs showed maximum sensitivity to modulation between 1 and 3 cycles/octave with reduced sensitivity above and below this region. Masked spectral modulation detection thresholds were measured for masker modulation frequencies of 1, 3, and 5 cycles/octave with a fixed modulation depth of 15 dB. The masking patterns obtained for each masker frequency and carrier band revealed tuning (maximum masking) near the masker frequency, which is consistent with the theory that spectral envelope perception is governed by a series of spectral modulation channels tuned to different spectral modulation frequencies.

Spectral modulation detection as a function of modulation frequency, carrier bandwidth, and carrier frequency region

The present study investigates the nature of spectral envelope perception using a spectral modulation detection task in which sinusoidal spectral modulation is superimposed upon a noise carrier. The principal goal of this study is to characterize spectral envelope perception in terms of the influence of modulation frequency (cycles/octave), carrier bandwidth (octaves), and carrier frequency region (defined by lower and upper cutoff frequencies in Hz). Spectral modulation detection thresholds measured as a function of spectral modulation frequency result in a spectral modulation transfer function (SMTF). The general form of the SMTF is bandpass in nature, with a minimum modulation detection threshold in the region between 2 to 4 cycles/octave. SMTFs are not strongly dependent on carrier bandwidth (ranging from 1 to 6 octaves) or carrier frequency region (ranging from 200 to 12 800 Hz), with the exception of carrier bands restricted to very low audio frequencies (e.g., 200-400 Hz). Spectral modulation detection thresholds do not depend on the presence of random level variations or random modulation phase across intervals. The SMTFs reported here and associated excitation pattern computations are considered in terms of a linear systems approach to spectral envelope perception and potential underlying mechanisms for the perception of spectral features.