Journal Logo

Journal Club

Frequency Compression: New Research Yields Clues for Patient Selection

Chung, King PhD

Author Information
doi: 10.1097/01.HJ.0000442743.11605.80
  • Free
© Shutterstock/FusionMS

Frequency-lowering strategies, which present high-frequency sounds to lower-frequency regions, have been used in audio recording and entertainment for a long time—e.g., changing the key of a karaoke song to suit the vocal range of the singer—and explored in hearing aids since the 1960s. Now, two new studies investigate the limits of the approach, particularly for those with lower cognitive function.

The rationale for frequency lowering is that many hearing aids cannot make sounds audible to their users because of the high-frequency roll-off of receivers, and individuals with high-frequency hearing loss or cochlear dead regions may not be able to make use of high-frequency sounds even if the sounds are audible.

While many different frequency-lowering strategies for hearing aid use have been proposed over the years, there are common features; namely, that sounds in the low-frequency region remain true to their frequencies, and only sounds in the high-frequency region are processed.

The differences between the strategies include, but are not limited to:

  • Whether the frequency-lowered sounds have unique representation in a frequency region or are superimposed onto the lower-frequency regions with other naturally occurring sounds.
  • Whether all high-frequency sounds are lowered, or just certain sounds with predetermined features.
  • Whether only the frequency-lowered sounds are presented at the output, or the frequency-lowered sounds are presented along with the original sounds at the output.
  • The settings of other parameters, such as the frequency-lowering threshold and the amount of lowering in the resultant signal.

Frequency-lowering strategies implemented in commercially available hearing aids generally can be divided into three categories:

  1. Frequency compression: High-frequency sounds (e.g., 4,000-8,000 Hz) are squeezed into a smaller frequency range (e.g., 4,000-6,000 Hz) above a certain frequency, called a frequency compression threshold (4,000 Hz in this example).
  2. Frequency compression is analogous to amplitude compression, whereby sounds with a large range of amplitude variations are packed into a smaller amplitude range. The frequency compression threshold is analogous to the kneepoint or compression threshold in the amplitude compression algorithm, in that frequencies above the threshold are compressed in both cases.
  3. Frequency transposition: All high-frequency sounds above the frequency-lowering threshold (e.g., 4,000-8,000 Hz) are superimposed onto sounds at a lower-frequency region (e.g., 2,000-6,000 Hz), but the spectral separation of high-frequency sounds is maintained.
  4. In this example, the 2,000 Hz to 4,000 Hz region contains both the naturally occurring sounds and the superimposed (lowered) sounds.
  5. Frequency translation: Whenever a high-frequency sound with predetermined features (e.g., s = 4,000-8000 Hz) is detected, the algorithm presents the sound at its natural frequency and generates a lower frequency sound (2,500-6,500 Hz), which is superimposed onto a lower-frequency region.
  6. The paradigm is coined frequency translation because a secondary s sound is “translated” down to a lower-frequency region.

Exploring the Limits of Frequency Lowering

Souza PE, Arehart KH, et al J Speech Lang Hear Res2013;56(5):1349-1363

Pamela E. Souza and colleagues examined the effects of a custom frequency compression algorithm (not commercially available) on speech recognition and perceived sound quality in adults with mild to moderate sensorineural hearing loss.

Institute of Electrical and Electronics Engineers (IEEE) sentences processed with frequency compression ratios of 0, 1.5, 2.0, and 3.0, and frequency compression thresholds of 1,000, 1,500, and 2,000 Hz, were presented to 26 listeners age 60 to 92 at signal-to-noise ratios of quiet (no noise), -10, -5, 0, 5, and 10 dB, for a total of 60 conditions ([3 frequency compression ratios x 3 frequency compression thresholds + 1 original condition] x 6 signal-to-noise ratios).

The speech level was fixed at 65 dB SPL, and the background noise level was varied. Listeners repeated the sentences, and they rated sound quality using an 11-point scale.

Speech recognition scores were reduced when a high frequency compression ratio with a low frequency compression threshold was used, whereas moderate frequency compression (i.e., 2.0) plus a high frequency compression threshold (i.e., 2,000 Hz) had minimal effects on sentence intelligibility, the results showed.

Regarding sound quality, listeners gave conditions with background noise lower ratings than they did quiet conditions. Participants also appeared to be more sensitive to frequency compression in quiet than in noise, such that increasing the amount of frequency compression (i.e., increase ratio or reduce threshold) would result in lower sound quality ratings in quiet, but not so in noise.

In addition, the sound quality degradation in quiet was more prominent for listeners with better high-frequency hearing.

Frequency compression can be a viable intervention for listeners with more severe high-frequency hearing loss, but not for all hearing aid users, the authors suggested. They also warned of a possible trade-off between improved audibility and increased distortion when fitting frequency compression hearing aids.

Working Memory, Age, and Hearing Loss: Susceptibility to Hearing Aid Distortion

Arehart KH, Souza P, Baca R, Kates JM Ear Hear 2013;34(3):251-260

The discussion of the paper on the limits of frequency lowering sets the stage for the analysis of a concurrently published paper by the same group of authors about working memory, age, and hearing loss.

In the study by Kathryn H. Arehart and colleagues, the working memory of (almost) the same group of listeners was tested. The researchers used a hierarchical linear modeling paradigm to analyze the factors contributing to the speech recognition scores the participants obtained in the earlier paper by Souza et al.

Instead of taking the sound-quality ratings from the earlier study, Arehart et al used the Hearing Aid Speech Quality Index (HASQI; Ear Hear 2010;31[3]:420-436; J Audio Eng Soc 2010;58[5]:363-381,_Nonlinear_Processing,_and_Linear.15.aspx) to estimate the amount of signal distortion introduced by frequency compression and the presence of background noise.

Working memory, age, and threshold at 4,000 Hz accounted for 29.3 percent, 11.5 percent, and 6.7 percent, respectively, of the variability in speech-recognition scores, their results indicated.

Figure. D
Figure. D:
r. King Chung, PhD

In addition, listeners with higher working memory appeared to be less susceptible to distortions created by frequency compression and background noise than those with lower working memory.

Clinicians should consider the possible degradation in hearing aid performance introduced by various forms of signal distortion when fitting hearing aids to older adults, especially those with lower cognitive function, the authors recommended.

© 2014 by Lippincott Williams & Wilkins, Inc.