Page 35 - Acoustics Today Summer 2011

quiet sounds to be made audible without making loud sounds uncomfortable. For patients suffering the reduced dynamic range that is typical of sensorineural hearing loss, compression can provide audibility and comfort over a wider
11
the overall signal level, and applied equally across all fre- quencies. This technique preserves the spectral shape of the processed signal, but it has the disadvantage that the gain computation is dominated by the region of the spectrum hav- ing the greatest energy. The presence of a strong, narrowband signal in one frequency region can thereby cause weaker sig- nals at distant frequencies to be rendered inaudible. Moreover, most patients suffer hearing loss that is non-uni- form in frequency. Wideband compressors offer no means of prescribing more compression in frequency regions of greater hearing loss.
In multiband compression, the signal is filtered into sev- eral frequency bands, by means of a filterbank or a discrete Fourier transform, and compression is applied independent- ly to the signal in each band. At any one time, the gain applied in a multiband compression prevents a strong, nar- rowband signal from triggering a gain reduction at distant frequencies, and allows compression to be prescribed differ- ently in each band according to the patient’s hearing loss. As many as 32 bands of compression may be employed in a hearing aid, depending on the manufacturer.
Compressive amplification is described by the ratio of input level (in decibels) to output level. A compression ratio
dynamic range than linear amplification.
In wideband compression, gain is computed according to
of 2:1, for example, implies that a change in input level of 2 dB produces a 1 dB change in output level. Compression ratios in hearing aids rarely exceed 3:1. Typically, sounds quieter than the compression threshold receive linear amplifi- cation (1:1), to avoid excessive amplification of low level background noise, and in some cases, they may be attenuat- ed (this is called dynamic range expansion).
Instantaneous gain changes introduce artifacts and compromise sound quality, so compression circuits are fur- ther characterized by a pair of time constants that determine how quickly the gain is reduced when a sudden increase in signal level is detected (the attack time constant) and how quickly the gain is restored due to a drop in signal level (the release time constant). Attack time constants are often short, on the order of tens of milliseconds or less, to prevent a sud- den loud sound being presented with painfully high gain to the hearing aid wearer. Release time constants are typically tens to hundreds of milliseconds. Time constants are often uniform across all bands, but need not be so.
While many people with hearing loss experience loud sounds similarly to people with normal hearing, a further consequence of hearing loss for some patients is an increased sensitivity to loud sounds, called hyperacusis. To prevent very loud sounds from causing discomfort or saturation, hearing aids may employ a further stage of heavy output limiting compression to keep the output signal within “safe” limits.
Multiband compression is the core of modern digital hearing aid signal processing, and is the primary tool for restoring audibility and comfort to patients with hearing loss.
In addition, many other signal processing algorithms are employed to increase patient comfort, to improve sound quali- ty, to provide microphone directionality, or to treat special conditions of hearing loss. For example, dedicated signal pro- cessing is applied to the problem of restoring audibility of high-frequency speech cues to patients with severe to profound high frequency hearing loss.
Frequency translation
High frequency sounds are critical to speech intelligibility, with a substantial portion of audible speech cues occurring at frequencies higher than 3 kHz. The highest frequency speech sound, the fricative /s/, is one of the most common consonant sounds in the English lan- guage, and its energy typically peaks above 5 kHz. For some patients with high-frequency, sloping hearing loss, restoration of audibility for these high- frequency speech cues may not be possi- ble with conventional amplification. Restoration of audibility for these patients is often constrained by the power avail- able in the hearing aid, by the amount of gain that can be applied without intro-
Fig. 6. Frequency transposition moves a high frequency spectral peak to a lower frequency within the patient’s range of aidable hearing. Vertical stems represent a set of harmonic frequency components, with the transposed components depicted in red. The unprocessed spectral envelope is represented by a black dashed line.
Speech and Hearing 31