Patients with a precipitous high-frequency hearing loss pose a unique challenge for dispensing audiologists. Hearing aids rarely provide sufficient high-frequency gain without over-amplifying the lower frequencies, which results in poor sound quality or significant sound distortions. The occlusion effect and upward spread of masking may also occur. For these patients, the high frequencies are not easily reachable.
Recently it has been suggested that a precipitous loss may make the high frequencies non-functional, "dead," or unaidable. Some studies suggest a decrease in word recognition scores when these "dead" regions are amplified. This result casts doubt on the usefulness of amplifying such a region. For people with either an unaidable or unreachable high-frequency hearing loss, information carried in those frequency regions would not be available despite the use of amplification.
What Is Frequency Lowering?
Audibility of unaidable or unreachable high-frequency information is achieved by converting that information to the lower frequencies where hearing is aidable or reachable. This conversion process is called frequency lowering. For example, information carried at 4000 Hz may be "lowered" to 2000 Hz so that it is heard as 2000 Hz. It is important to recognize that frequency lowering doesn't restore hearing in the original high frequencies. Instead, the signals are heard at a lower frequency. For this reason, a frequency-lowered signal will initially sound "unnatural" to most listeners.
The concept of frequency lowering is not new. Braida, Durlach, Lippmann, Hicks, Rabiowitz, and Reed (1979), in an excellent review of frequency-lowering techniques, dated studies on the topic beginning in the 1950s. Early attempts at frequency lowering implemented on analog technologies made it difficult to achieve the desired signal-processing results without introducing significant artifacts, such as unnatural-sounding speech and distorted temporal and rhythmic patterns. The studies acknowledged the shortcomings, but proceeded without correcting for signal distortions.
With few exceptions, most studies reported no substantial improvement in speech recognition resulting from these early frequency-lowering techniques. In some cases, decreased performance over conventional amplification was noted. The researchers hypothesized that the early negative results were due to limitations in technology, lack of individualized fittings, and the subject's degree of hearing loss, as well as the lack of patient training and/or sufficient experience with the feature.
The introduction of digital signal processing (DSP) allowed researchers to achieve the desired signal processing results while minimizing the artifacts experienced by earlier approaches. With careful selection and training of patients, the recently introduced frequency-lowering algorithms in commercial hearing aids have shown significant benefits for adults and children with hearing loss (Auriemmo et al., 2008; Kuk, Peeters, Keenan, & Lau, 2007). These benefits include better aided high-frequency sound-field thresholds, improved perception of high-frequency sounds in nature such as birds singing or leaves rustling, improved identification of voiceless consonants, and increased accuracy in production of voiceless fricatives for children.
Types of Frequency Lowering
Frequency compression and frequency transposition are two major approaches to lowering the unaidable or unreachable high-frequency sounds. The approaches differ in how the original aidable sounds are affected and how much of the frequency spectrum is involved. In frequency compression the whole or a portion of the frequency spectrum is compressed to fit into a narrower frequency region. For example, if the wearer has no hearing above 4000 Hz, information across all frequencies will be compressed into the frequency region below 4000 Hz (i.e., the wearer's residual frequency range). If a compression ratio of 2 is used, an 8000 Hz tone will be compressed to a 4000 Hz tone, the original 4000 Hz tone will be compressed to 2000 Hz, etc. Although sounds in the previously dead region above 4000 Hz are now heard, frequencies in the aidable region from 0 Hz to 4000 Hz that do not need to be lowered are nonetheless compressed.
In contrast, frequency transposition moves the higher frequencies in the unaidable or unreachable region to a lower-frequency region where they are superimposed. This method spares the lower frequencies from unnecessary compresssion. Again, if the unaidable region is above 4000 Hz and the transposition targeted is one octave, 4000 Hz will be moved to 2000 Hz and 8000 Hz will be moved to 4000 Hz. However, the original 4000 Hz will remain at 4000 Hz, and 2000 Hz and 1000 Hz (and so forth) will remain at their original frequencies where they will be mixed with the transposed sounds above 4000 Hz. The advantage of this approach is that it preserves the temporal structure of the original signals, although it could result in masking the original sounds and cause initial confusion.
Current commercial products use each approach. Because each approach is unique, the efficacy of one type of frequency lowering cannot be generalized to another type. This article summarizes the efficacy data for one commercial frequency-transposition hearing aid.
Listening With Frequency Transposition
One important consideration in evaluating a frequency-lowering device is to include acoustic stimuli of various spectral complexities because the perceived benefit varies with the spectral complexity. In a study (Kuk et al., 2007) of preference in using frequency transposition over the non-transposition when listening to bird songs, music, and female conversation, the preference for frequency transposition was the highest when listening to bird songs (more than 90%), followed by music (70%), and speech (60%) (see Figure A [PDF]).
This relative preference is easy to understand. Without transposition, bird songs (which have a narrow frequency range) would not be audible because of the subjects' high-frequency hearing loss. With transposition, what is previously inaudible becomes audible at a lower frequency. For music and speech, the stimuli contain a wider range of frequencies that are partially audible to the listeners without transposition. Consequently, the contribution of transposition is less evident. The addition of the transposed sounds to the original sounds could also lead to an initial "unnatural" perception for some wearers.
Evaluating frequency transposition with everyday and nature sounds becomes more relevant when hearing loss is in the severe-to-profound range. These patients may value the ability to hear environmental sounds for safety and comfort, even though their ability to perceive these sounds may not reflect their ability to understand conversational speech.
For the right patients, frequency transposition may improve speech intelligibility in comparison to conventional amplification. Kuk, Peeters, Keenan, and Lau (2007) evaluated the efficacy of a transposition algorithm on 13 adults with a sloping or precipitously sloping high-frequency hearing loss. After using frequency transposition for one month, the average adult improved consonant identification by 10%–15% over the no-transposition program at a soft input level (30 dB HL) and by 5%–10% at a conversational level (see Figure B [PDF]).
A study by Auriemmo et al. (2008) showed improvement in consonant intelligibility for children aged 6¾ and 12 years old. At the end of the six-week trial with frequency transposition, the average child improved consonant identification by 20% at a 30 dB HL input level (from 48% to 68%) and 5%–10% at a 50 dB HL input level (Figure C [PDF]).
Both studies evaluated speech perception using the Nonsense Syllable Test, which is scored on a phoneme basis. The advantage of this approach is the ability to analyze the types of errors that wearers make. In both the adult and pediatric studies, voiceless consonants such as /s/, /sh/, /ch/, /th/, /p/, and /t/ were identified more accurately with transposition than without transposition after an adjustment period. In general, perception of voiceless consonants, especially stops and fricative sounds, is more likely to improve with transposition. Because of subjects' unfamiliarity with transposed signals, however, improvements may not be evident initially, but experience with frequency transposition may facilitate correct sound identification.
Improving Children's Speech
Better hearing is frequently reported to improve speech and language skills in children. We compared the consistency and accuracy of the production of /s/ and /z/ for 10 children during a five-minute, open-ended storytelling task and reading task when they wore their own digital hearing aids, the study hearing aids without the transposition (master), and the study hearing aids with transposition (AE). These children had a moderate-to-severe hearing loss in the low-to-middle frequencies and a severe-to-profound loss above 2000 Hz. The children's speech production was recorded and later analyzed by a speech-language pathologist.
The number of times the children produced the /s/ and /z/ phonemes correctly was counted and compared to the number of times they could have produced these sounds. Figure 1 [PDF] shows the accuracy of the production at the baseline (with own aid), with the master program (no transposition) at three weeks, and with the transposition program after six weeks. The average accuracy score was approximately 70% with the children's own hearing aids, 80% with the master hearing aids after three weeks, and almost 90% after using the transposition program for six weeks. This difference was statistically significant (p < 0.05).
Prior to the study, SLPs working with several of the children discharged them because of a lack of improvement with intervention. Following the study, these children were re-enrolled in speech-language treatment because the SLPs felt the children could now improve their speech with the additional audibility provided by the transposition algorithm.
Frequency transposition is not for everyone with a hearing loss. If candidates are selected inappropriately and/or the device is fit improperly, frequency transposition could lead to speech recognition that is poorer than that provided by traditional amplification. This algorithm is intended only for those whose hearing loss cannot be helped with traditional amplification because of unreachable gain or a cochlear-dead region. Typically, people who have a profound loss in the high frequencies along with a moderate loss in the low-to-middle frequencies are excellent candidates; those who have a precipitous hearing loss in the high frequencies are also good candidates.
Frequency transposition is beneficial when the hearing loss above 3000 Hz is 70 dB HL (see Figure D [PDF]). The greater the hearing loss is in the high frequencies, the greater the potential benefit in consonant identification with frequency transposition. However, hearing loss below 2000 Hz should be aidable, but not too severe so that there is still good temporal and frequency resolution of those hair cells. Thus, the hearing loss in those regions can range from normal hearing to as much as a moderate hearing loss. The hearing loss of candidates should not be more than 70 dB HL at 1000 Hz and 90 dB HL at 2000 Hz.
The selection of a transposition feature should also be based on the listening needs of listeners with hearing loss. Although the ability to hear missing high-frequency speech sounds is an expected need for most candidates, the need to hear nature sounds (such as bird chirping) and everyday sounds such as the timer on the oven or microwave is also important, even though the wearers' hearing loss may be outside the fitting range shown in Figure D [PDF]. Precise identification of the best candidates will be possible as we gain more experience with this technology.
Need for Training
An adjustment period is usually required before the full benefit of frequency transposition is realized, especially for listening to speech and other complex signals. Figure 2 [PDF] shows the consonant scores over time for the 10 children evaluated in our study using frequency transposition. The children received weekly half-hour auditory training with a three-week interval between sessions. Their consonant identification scores increased over time, with a difference of as much as 20% between the initial score and the score after six weeks of auditory training and use of frequency transposition. Children may take less time to adjust to the algorithm and continued improvement may be observed up to three months after initial use. For adults, it may take one to two months to fully adjust to the sound quality of the transposed speech signal and realize its benefits in aiding speech discrimination. Auditory training may facilitate the wearer's attention to the transposed sounds and help users adapt more quickly to frequency transposition.
These studies used small subject samples and limited sources, and the conclusions about the efficacy of frequency transposition must be viewed as tentative. The data are encouraging because they represent an evolution of research based on new technology (e.g., Braida et al., 1979). Early attempts at frequency lowering implemented on analog technology reported no substantial improvement in speech recognition, but with the introduction of digital signal processing and careful patient selection, we are now achieving the desired signal-processing results.
We believe these findings will pique renewed interest among audiologists and researchers in conducting independent research on the efficacy of frequency transposition. Care must be taken in selecting appropriate candidates for the technology and in providing appropriate fitting and counseling/training so they may realize the full benefit of frequency transposition.
Disclosure: The primary research described in this article was funded by the authors' employer, a hearing aid manufacturer.