TI数字助听器设计
Digital Hearing Aids
Digital Hearing Aid design considerations and challenges, and suggested products.
Design Considerations
Design Challenge
Designers of hearing aids have stringent technological requirements. Hearing aids must be small enough to fit inside or behind ones ear, run with extremely low power, and introduce no noise or distortion. To achieve these requirements, current hearing aid devices consume less than 1 mA, operate at 1 V, and utilize less than 10 mm2 of silicon area which usually means two or three devices stacked on top of each other. The typical analog hearing aid consists of an amplifier with a non-linear input/output function and a frequency dependent gain. However, this analog processing suffers from a dependency on custom circuits, lack of programmability and a higher cost when compared to digital processing. Recent digital devices have reduced device costs and lowered power consumption compared to their analog counterparts. The greatest advantage offered by digital devices is their improved processing power and programmability, allowing hearing aids to be customized to a specific hearing impairment and environment. Instead of a simple sound amplification and adjustable frequency compensation, more complex processing strategies can be achieved to improve the sound quality presented to the impaired ear. Such strategies, however, require the very sophisticated processing that a DSP can provide.
Typically, hearing loss is divided into two categories: conductive hearing loss and sensorineural hearing loss (SNHL). A conductive hearing loss occurs when the transduction of sound through the patient´s outer or middle ear is abnormal, and a sensorineural hearing loss occurs when either the sensory cells in the cochlea or the neural mechanisms higher in the auditory system fail.
With a conductive hearing loss, sound is not properly transmitted through the middle or outer ear. Because sound is primarily attenuated with a conductive loss, amplification of sound is essentially all that is required to restore near-normal hearing. No special signal processing is necessary, and traditional analog hearing aids work well. However, only 5% of those inflicted with some hearing loss are attributed to conductive losses alone.
The other type of hearing loss is SNHL. This includes hearing loss that is associated with aging, as well as noise-induced hearing loss and loss caused by drugs that are harmful to the auditory system. Most types of SNHL appear to be caused by a cochlear malfunction. SNHL is thought to be caused by damage to inner hair cells, outer hair cells, or both. However, the underlying physiology is complicated. Different people will have different pathologies which means that patients with identical audiograms will not necessarily have the same kind of hearing loss. Further, patients may even have differing kinds of impairment over different frequency ranges.
The effects of SNHL usually result in: 1) lack of input in some frequency channels, 2) lack of sensitivity, and 3) widened auditory filters. These effects, in turn, significantly impact the listeners perception of sound. Compared to normal hearing listeners, listeners with SNHL will most likely experience loudness recruitment (the range of comfortable listening levels is compressed when compared with normal) and loss of frequency resolution, among other difficulties. These changes in sound perception have significant effects on a listeners ability to understand speech.
Because SNHL is not simply a problem with the transmission of sound, but actually a problem with the processing of sound, this loss is not likely remedied through simple amplification - making garbled sounds louder does not make them clearer. Therefore, one potentially effective way to help an SNHL patient is through pre-processing the signal to enhance complex tonal patterns to compensate for the hearing loss.
It is unlikely that the various manifestations of SNHL will be remedied by the same optimal treatment. Processing of the sound can make speech more intelligible. However, the best processing algorithms will differ among individuals and may even change for one individual in different listening conditions such as a quiet room versus a noisy stadium. The key to accommodating these differences is hearing aid flexibility.
Traditionally, hearing aids have been amplifiers encased in custom earmolds fitted to the end user. The hearing aid system contains a microphone, an amplifier, a Zinc-Air battery and a receiver/speaker. Most of these amplifiers incorporate some kind of compression function, essentially a non-linear input/output relationship, that is used to compensate for loudness recruitment. Also, the gain in different frequency bands can be adjusted, and the number of frequency bands varies, but is usually two or three bands. Many of the newest aids are digitally programmable, which means that although they have analog signal processing, the processing is controlled by digital parameters that can be adjusted by an audiologist. In addition, some analog aids have several programs, or sets of parameters, for different listening environments.
Some of the digital hearing aids in the market are ASICs with programmable coefficients. These ASICs provide a few sets of algorithms and several frequency bands that were not possible in typical analog devices. For example, the digital hearing aids have a combination of the following features: 2 to 14 frequency bands with adjustable crossover frequencies, one microphone, dual microphones for directional listening, background noise reduction, automatic gain control (AGC), speech enhancement, feedback reduction and loud sound protection. Overall, the amount of processing that can be done is impressive, especially when compared with the traditional processing in an analog aid.
Design Example
A DSP-based hearing aid could expand software-controlled features to include: frequency shaping, feedback reduction, noise reduction, binaural processing, pinna and ear canal filtering, reverberation reduction and a provision for direct digital input from a digital telephone, TV or other audio devices. A programmable DSP also means that the hearing-aid algorithms/features could be customized or changed without changing the hardware. Hearing-aid practitioners could economically experiment with available algorithms on a near real-time basis. It would even be possible to have user-selectable programs for switching to highly processed sound in difficult listening situations or back to traditional, less distorted sound in quiet environments.
DSP-based Hearing Aid Block Diagram
The block diagram above shows the primary elements for a DSP-based digital hearing aid. A typical digital hearing aid consists of three semiconductor die stacked on top of each other: EEPROM or non-volatile memory, a digital device and an analog device. Recent advances have allowed the integration of these modules into two or even one semiconductor die. Due to the battery´s range of voltage from 1.35 V to 0.9 V, these devices are designed to operate at 0.9 V. Some implementations use power management to monitor battery voltage and alert the user when the battery is low and gracefully shutdown the system when the voltage drops too low. The analog device normally includes the sigma-delta analog-to-digital converter, microphone preamplifier with compression input limiting function, remote control data decoder, clock oscillator and voltage regulator. The sigma-delta A/D typically has a frequency range of 20 kHz with 16-bits of resolution (14-bit linear). The digital device includes the DSP, logic support functions, programming interface and the output stage. The output stage is normally all digital and uses a pulse-width-modulated (PWM) output with Class D amplifiers that utilizes the speaker impedance to perform the analog-to-digital conversion.
Overall, the power consumption in current analog and digital hearing aids are approximately equal. Total current consumption is about 0.7 mA to 1.0 mA in the analog devices, whereas digital devices consume 0.5 mA to 0.7 mA. A 1.35-V zinc-air battery that provides around 30 to 65 mAh with a 50-µA self-discharge current powers this system. The end-of-life voltage is about 0.9 V. Due to the increased amount of processing in the digital aids, however, a straight comparison of consumption between digital and analog aids is not entirely fair. Digital aids with processing abilities equivalent to those of analog aids would consume even less power.
Featured Products
AIC111 - Micropower Audio Codec
The AIC111 is a Micropower DSP or microcontroller-compatible audio codec that provides a high-performance analog interface solution for applications such as personal medical devices - hearing aids, aural preprocessing - and low-power headsets. The AIC111 supports a 1.3-V CMOS digital SPI interface, and includes an external microphone supply and bias, and low battery monitor and indicator