Effects of sensorineural hearing loss on temporal coding of narrowband and broadband signals in the auditory periphery

Abstract

People with sensorineural hearing loss have substantial difficulty understanding speech under degraded listening conditions. Behavioral studies suggest that this difficulty may be caused by changes in auditory processing of the rapidly-varying temporal fine structure (TFS) of acoustic signals. In this paper, we review the presently known effects of sensorineural hearing loss on processing of TFS and slower envelope modulations in the peripheral auditory system of mammals. Cochlear damage has relatively subtle effects on phase locking by auditory-nerve fibers to the temporal structure of narrowband signals under quiet conditions. In background noise, however, sensorineural loss does substantially reduce phase locking to the TFS of pure-tone stimuli. For auditory processing of broadband stimuli, sensorineural hearing loss has been shown to severely alter the neural representation of temporal information along the tonotopic axis of the cochlea. Notably, auditory-nerve fibers innervating the high-frequency part of the cochlea grow increasingly responsive to low-frequency TFS information and less responsive to temporal information near their characteristic frequency (CF). Cochlear damage also increases the correlation of the response to TFS across fibers of varying CF, decreases the traveling-wave delay between TFS responses of fibers with different CFs, and can increase the range of temporal modulation frequencies encoded in the periphery for broadband sounds. Weaker neural coding of temporal structure in background noise and degraded coding of broadband signals along the tonotopic axis of the cochlea are expected to contribute considerably to speech perception problems in people with sensorineural hearing loss. This article is part of a Special Issue entitled "Annual Reviews 2013".

Keywords: AN; CF; ENV; FM; SPL; SR; TFS; auditory nerve; characteristic frequency; envelope; frequency modulation; sound pressure level; spontaneous rate; temporal fine structure.

Fig. 1

Schematic of the spectral decomposition of broadband sounds and coding of TFS and ENV information in the cochlea. Broadband acoustic input signal (top) and output signals (left) of auditory filters with center frequencies between 1.5 and 5 kHz. Each output signal consists of a slowly varying amplitude envelope (ENV; red) and a quickly varying temporal fine structure (TFS, black). Auditory filter output signals are encoded in the spike rate of auditory-nerve fibers (right). In fibers with characteristic frequencies (CFs) below 4–5 kHz, spike rate varies with both the TFS and ENV of the auditory filter output signal. In fibers with higher CFs, spike rate varies primarily with the ENV of the output signal, due to the low-pass membrane filtering of the inner hair cells.

Fig. 2

Phase locking to the TFS of pure tones in quiet and in background noise. Vector strength of phase locking to pure tones in chinchilla auditory-nerve fibers is plotted as a function of characteristic frequency. Control and noise-exposed fibers show similar phase locking to tones under quiet conditions. A significant reduction in the vector strength of phase locking in the noise-exposed population emerges under noisy conditions. Noise level is the root-mean-square amplitude of the noise expressed in dB relative to the root-mean-square amplitude of the tone. (Modified from Henry and Heinz, 2012, and originally published in Nature Neuroscience by Nature Publishing Group).

Fig. 3

Post-stimulus time histograms of auditory-nerve fiber responses to 50-ms tones presented at characteristic frequency (CF). Fibers from chinchillas with noise-induced hearing loss show greater onset response amplitude (left-pointing arrows), reduced onset latency, faster adaptation during stimulation, and slower recovery following stimulus offset. Black and gray lines show mean responses of normal-hearing fibers and noise-exposed fibers, respectively, for populations of neurons with CFs between 1 and 4 kHz. Separate panels present data from fibers falling into different categories of spontaneous firing rate (SR): low (SR<1), medium (118). Modified and reprinted from Scheidt et al. (2010), with permission from Elsevier.

Fig. 4

Loss of synchrony capture of vowel formants with noise-induced hearing loss. Representative neural responses in cats to the vowel /ε/ (top) from a control auditory-nerve fiber (middle plots) and noise-exposed fiber (lower plots). The characteristic frequencies (CFs) of the fibers were near F₃. While the control fiber shows strong synchrony capture of TFS information near F₃ (i.e., near CF), the noise-exposed fiber shows synchrony to a broader range of lower frequency TFS information. Modified and reprinted from Miller et al. 1997, with permission from the Acoustical Society of America.

Fig. 5

Effect of cochlear hearing loss on the frequency tuning of temporal responses to TFS and ENV information. Representative frequency tuning curves (top) of chinchilla auditory-nerve fibers with varying degrees of noise-induced hearing loss. Characteristic frequencies (CFs) ranged from 3 to 4 kHz. Wiener kernels (middle panels) show the temporal response to TFS information (1^st order kernels) and ENV information (filtered 2^nd order kernels). Frequency-domain representations of the kernels (bottom panels) show that control fibers in this CF range encode primarily ENV information (red) based on stimulus energy near CF. Mild hearing loss introduced additional phase locking to low-frequency TFS information (black curve), while with moderate impairment both the TFS and ENV information encoded were centered at lower frequencies well below CF.

Fig. 6

Derivation of temporal-modulation transfer functions from second-order Wiener kernels. Spectro-temporal receptive fields (top), modulation tuning functions (middle), and temporal-modulation transfer functions (bottom) for the control fiber and moderately impaired fiber shown in Fig. 5 (CFs between 3 to 4 kHz). Noise floors (dotted lines in bottom panel) were calculated from the temporal regions before (0 to 2 ms) and after (last 3 ms, e.g., 7 to 10 ms) the response in the spectro-temporal receptive fields. Noise-induced hearing loss enhanced the coding of faster temporal modulations.