Dynamic range compression

Dynamic range compression is a method which normalizes the dynamic range of hearing aid output between a listener’s auditory thresholds (minimum audible ouput) and uncomfortable listening levels (maximum allowable output). It does so by making level-

dependent gain adjustments as the level of speech entering the hearing aid fluctuates. Typically, a hearing aid applies compression by amplifying soft passages more than loud passages to increase overall audibility. However, there is some flexibility to how compression is applied. Parameters (further described below) including compression ratio, time constants, and number of frequency channels can all be manipulated to apply compression differently. The application of minimal compression ratios and longer time constants is typically ideal for optimal speech quality (Souza, 2002). It is of interest to determine if similar parameter recommendations can be made for music quality.

The compression ratio is the parameter which determines how much gain is applied at each level. A compression ratio of 2:1 would imply that for every doubling of signal level, half the gain would be applied at the higher level. This concept contrasts linear gain (1:1), in which constant gain is applied to the signal regardless of its level, and expansion (0.5:1), in which less gain is applied to very soft inputs to reduce the amplification of processing noise. Several studies have examined the effect of manipulating compression ratio and ratings of amplified music. For example, van Buuren, Festen, & Houtgast (1999) investigated the effect of various compression ratios (0.25:1, 0.5:1, 1:1, 2:1, and 4:1) for several music genres. They found that the highest pleasantness ratings were associated with the reference linear amplification (1:1) condition across all genres. Subsequent studies showed a similar trend. When listening to a variety of music genres processed using multiple compression ratios, linear or linear-like gains were typically preferred relative to minimal compression ratios, which were in turn preferred relative to larger compression ratios (Arehart et al., 2011; Croghan, Arehart, & Kates, 2014; Higgins, Searchfield, & Coad, 2012; Kirchberger & Russo, 2016a). In addition, listeners reported greater clarity of individual instruments when listening using linear amplification compared to wide- dynamic range compression (Madsen, Stone, McKinney, Fitz, & Moore, 2015). It is, however, possible that linear gain can be problematic if it cannot reproduce the signal of interest with fidelity. Exceptionally loud signal peaks can exceed the dynamic range of the analog-to-digital converter of the hearing aid, which can introduce distortions caused by peak-clipping and output limiting. In one study, hearing-impaired listeners rated the quality of music amplified using wide-dynamic range compression or linear amplification with peak-clipping or output limiting. Listeners slightly preferred the compressive settings to

the other settings (Davies-Venn, Souza, & Fabry, 2007). Clinicians fitting hearing aids should be wary of the level of the signal being amplified by a hearing aid and if distortions produced by compression are more detrimental to sound quality compared to peak-clipping or output limiting.

Compression time constants refer to the speed it takes for compression to be applied in response to a level change. When listening to a stimulus with a fixed average level, the level may fluctuate from moment to moment leading to a corresponding change in gain from moment to moment. Attack times are the time required for the gain to adjust to a new signal level, and release times are the time required for the gain to return to its default setting. Large, rapid changes in gain can lead to distortions, which may be disruptive to the listener. When listening to amplified music with release times of 40 ms or 4 seconds, hearing-impaired listeners consistently preferred the longer release time relative to the shorter (Hansen, 2002). However, there were also some genre and level interactions. Moore et al. (2011) identified that slow time constants (50 ms attack time, 3000 ms release time) were preferred to faster time constants for classical and jazz music at 80 dB SPL, while slow time constants were preferred to faster time constants for classical music at 65 dB SPL. In another study, slow time constants (50 ms attack time, 1000 ms release time) were preferred to fast time constants (5/50 ms) for both classical and rock music (Croghan et al., 2014). However, when comparing compressed music with time constants between 10 ms, 70 ms and 200 ms, there were essentially no preferences (Arehart et al., 2011). Together these results encourage the use of long time constants that have an attack and release time of at least 50 ms and 1000 ms, respectively, for optimal music sound quality.

Finally, compression can be manipulated on a channel by channel basis. Multiple channels can allow for more compression at frequencies where listeners may have elevated thresholds relative to other frequencies. While this can help practitioners have greater control of the hearing aid signal-processing, it also has the potential to disrupt spectral peak-to-valley differences which provide balance to the musical spectrum. In general, a single channel, or multiple channels with similar compression ratios, has been recommended for a balanced musical spectrum (Chasin, 2006). When hearing-impaired listeners rated sound quality of music processed by 1, 4 or 16 processing bands, listeners

preferred a single band most frequently (van Buuren et al., 1999). Furthermore, fewer channels (3 vs 18) were preferred for rock music, but not classical music (Croghan et al., 2014). However, in both studies, whether fewer channels were preferred was also a function of whether there was a notably high compression or shorter time constants. Therefore, when considering the number of channels in a compressive hearing aid, it is important to consider it in conjunction with other factors such as compression ratio, time constants, and stimulus.