Microsoft word - noisetrackerii 090723.doc

Abstract The main complaint of hearing instrument wearers continues to be hearing in noise. While directional microphone technology exploits spatial separation of signal from noise to improve listening in noise, a challenge for single microphone noise reduction systems is that the signal of interest is embedded in the background noise. ReSound NoiseTracker™ II uses a sophisticated algorithm to reduce amplification of noise for both single and dual microphone devices without impacting speech understanding. It accurately identifies speech and characterizes noise, operating seamlessly to reduce noise even during the pauses in running speech. In the clinic, ReSound NoiseTracker™ II can be customized for different listening preferences and/or listening situations. Unwanted amplification of background noise is a usage and satisfaction with amplification. Nabelek significant factor influencing hearing instrument and her colleagues proposed that hearing user satisfaction (Kochkin, 2000). Just as instrument usage might be determined more by directional microphone technology has risen wearers’ willingness to listen in background noise enormously in popularity as a means to combat than how well they understand speech (Nabelek et hearing-in-noise issues, noise reduction algorithms al, 1991). Using the acceptable noise level (ANL) have become an ever-present feature in digital test with elderly listeners, they found that full-time background noise than did part-time users or users This paper reviews the standard methods of single who had rejected their hearing instruments microphone noise reduction and describes the (Nabelek et al, 2006). These findings would ReSound NoiseTracker II system, an important suggest that eliminating or reducing background component of the surround sound by ReSound. noise when listening with hearing instruments NoiseTracker II is unique in its ability to reduce would increase the acceptance rate of hearing amplification of noise without affecting the instruments and provide significant benefit to a audibility or quality of simultaneously occurring speech. This provides not only increased listening Perhaps related to the willingness to listen in noise, listening effort in background noise also has far- Even though hearing instrument users of all ages instruments. Consider, for example, the hearing complain about background noise, it can be an instrument wearer who is attempting to listen to especially problematic issue for older adults, who speech in a noisy environment over a prolonged comprise the largest segment of users. The period of time. The concentration required to average hearing instrument user is 70 years old follow what is being said can be an exhausting (Kochkin 2005). In addition to hearing loss, many task. Conceivably, less effort would be required by older adults often experience problems with the listener if the background were less noisy, even auditory processing and cognitive function, which if overall understanding of speech was not can result in difficulty focusing, remembering or improved. Less effort required of the hearing processing information they have heard (Wingfield instrument user may even allow them to perform 2005). Studies have shown that even when younger dual attention tasks that normal listeners typically and older adults have similar hearing thresholds take for granted, such as listening to a talker while and speech discrimination ability in quiet, older also paying attention to their surroundings, adults will perform worse on speech discrimination simultaneously noting other conversations that may tasks in noise than their younger counterparts be going on around them, or monitoring the activity (Kricos 2006). In addition, Tun et al (2002) of other people in the room. Simply eliminating a suggested that higher level cognitive factors may distracting or annoying sound, such as a nearby air affect the ability of older adults to process language conditioner could, in other words, provide in the presence of competing speech. In terms of significant benefit to a hearing instrument user hearing instrument use, understanding speech in noise is likely to be particularly challenging for the older adult, even if audibility is provided. This in turn impacts success with amplification. Noise reduction and dual microphone directionality in hearing instruments were both conceived as a Apart from issues related to processing speech in means to improve speech understanding in noise. noise, other noise-related factors can influence Directionality achieves this goal by preferentially amplifying signals coming from a particular were based on signal modulation and are in fact location, such as in the look direction of the still the most common type of system today. hearing instrument user. For noise reduction to be Modulation-based noise reduction systems analyze effective in improving speech understanding in the level fluctuations of the input signal in different noise, it would have to separate desired speech frequency bands. Because single-talker speech can from the competing signals after both have been fluctuate in level more than 30dB, the assumption picked up by the same microphone. Even then, the is made that large fluctuations in the input sound hearing instrument would need to be capable of indicate good signal-to-noise (SNR) ratios, and that reducing amplification of only the noise components of the incoming sound without fluctuations decrease. If the SNR is estimated to be disturbing the speech components. This is a poor in a particular frequency band, gain is reduced daunting task, as both speech and what constitutes “noise” for the listener are similar in terms of frequency content. In fact, the undesired noise Although the underlying principle is the same, often is speech from multiple talkers. Not modulation-based systems differ in a number of surprisingly, the benefit of noise reduction has been important ways. One obvious difference is the noted more in its ability to improve listening amount of gain reduction. Some systems may allow comfort and effort than for improving speech more than 10dB reduction in gain while the understanding (Mueller et al, 2006; Bentler et al, maximum reduction for others can be 5dB or less. Another difference is that some algorithms may consider other acoustic properties of the input than Prior to digital hearing instruments, noise reduction just modulation in determining the amount of gain systems limited low frequency amplification by reduction. For example, they may reduce the gain means of a high pass filter or by means of different differentially depending on the rate of modulation compression strategies. The rationale behind this or on the overall input levels. Finally, systems approach was that noise – including multitalker differ in how long they take to decrease gain or babble - contains more energy at low frequencies restore it back to its original value in a given than the speech of a single talker, which could be channel. As in describing the dynamics of expected to contribute significantly to excessive compression systems, these are collectively called overall loudness of the signal. Also, low frequency the time constants. Systems which reduce and noise may cause upward spread of masking, restore the gain slowly can make listening in thereby making the high frequency parts of speech situations with stationary noise sources more comfortable, but may degrade the audibilility of speech. For this reason, most systems incorporate The advent of digital technology enabled more faster release times so that the gain is restored advanced means of determining the composition of quickly when a highly modulated signal such as the input signal. A better characterization of the speech is detected. Systems which also have fast sound entering the hearing instrument microphone attack times may negatively affect the overall could make it possible to limit gain reduction to sound quality as gain is rapidly increased and periods of time and spectral regions where important speech information was occurring. The first digital noise reduction algorithms to appear Figure 1. Steady-state noise is treated differently by different noise reduction systems. Figure 2. For babble noise, modulation-based noise reduction systems do not affect the background noise. NoiseTracker™ II is different than traditional modulation-based systems as it is able to reduce the level of the babble while preserving the level of speech. As a result of such differences, modulation-based applications. The concept of spectral subtraction, noise reduction systems also differ greatly in how illustrated in Figure 3, is to subtract the short-term various types of input sounds are treated (Bentler & noise spectrum from the total signal, leaving only Chiou, 2006). An example of this is illustrated in the speech portion. Although the concept is easy, Figure 1 which shows how different types of the implementation is not. The success of this system react to steady-state noise with a short strategy hinges on being able to identify speech and speech passage at the end of the timeframe. Figure to precisely characterize noise. An additional challenge is to keep up with the dynamic speech background noise and the same short speech and noise make-up of real listening environments. passage at the end of the timeframe. There are Finally, it is important for hearing instrument users clear differences in both reaction time as well as that not all noise be removed from the signal, and amount of gain reduction for the same input. While that the noise characteristics be preserved. If all no evidence exists supporting the superiority of any ambient noise were removed or if the spectrum of particular system, there is one apparent the noise background was altered, this would create shortcoming of the modulation-based approach. an unnatural-sounding experience. Background This is the inability of such a system to accurately sounds do need to be audible to the degree that identify when speech is present. As a result, users can recognize and orient themselves in their modulation-based systems are more likely to make listening environments. Ultimately, the goal is “mistakes” in terms of when and in what undistorted speech at the prescribed gain, and frequencies gain reduction would be beneficial. Like other noise reduction schemes, NoiseTracker II has the goal of suppressing noise in frequency regions where the speech-to-noise ratio is low. However, NoiseTracker II is distinguished from the modulation-based method by its ability to reduce unwanted noise from the incoming signal without appreciably affecting the speech portion of the signal. The NoiseTracker II system is able to accomplish this because of 1) the higher degree of accuracy with which it identifies speech and noise when compared to other systems, 2) an adaptive Figure 3. Spectral subtraction removes noise from noise estimate, 3) fast time constants, and 4) a the total signal, leaving the desired signal intact. mathematically optimal gain reduction function based on SNR rather than signal modulation. The NoiseTracker II signal analyzer itself is comprised of three main components: a signal Built on the ReSound Warp-based platform, the power tracker, a speech presence indicator, and a NoiseTracker II system uses spectral subtraction noise power tracker. These three components (Boll, 1979), one of the most widely used methods provide information that is used to estimate the for enhancement of noisy speech in audio signal-to-noise ratio (SNR), which in turn determines the amount of gain reduction to be employs time constants for each of its tracking components as well as for the actual gain reduction and restoration. While the signal power tracker The signal power tracker represents the overall always works quickly in order to preserve the signal including speech and noise, and is the part speech envelope, the noise power tracker from which any noise will ultimately be subtracted. adaptively adjusts its time constants depending on The speech presence indicator analyzes the acoustic whether speech is detected. Estimation of the noise characteristics of the signal at 1-millisecond spectrum is thus limited to pauses between words intervals to determine the probability that speech is and syllables. In this way, the system avoids present in the overall signal. More specifically, the mistaking speech for noise, and prevents speech speech presence indicator looks for a temporal- information being subtracted from the overall spectral pattern of alternating high and low input. Once noise is detected, the time it actually frequency sound which is typical of speech. This takes for the decrease in gain to begin is within method constitutes a more precise way of 2 seconds. As the SNR improves or decreases identifying speech than relying on modulation further, new gain calculations are effected almost alone. With such an accurate identification of speech, the noise power tracker is able to restrict analysis of the noise background to frequency regions and points in time where speech is not NoiseTracker II system offers flexibility in the mixed with noise. It is critical for the system that degree of noise reduction it offers to address the noise estimate not be contaminated by speech, individual user preferences. Up to four options are since this would lead to distortion of the speech available: mild (-3dB), moderate (-6dB), considerable (-8dB) or strong (-10dB) noise reduction. The noise reduction value for each Once the overall signal and noise information have degree of NoiseTracker II is the amount which been provided by the speech presence indicator and would be applied when the estimated SNR is 0dB the noise power tracker, the SNR ratio is estimated or worse. A lesser amount of gain reduction would by comparing the level of the noise with the level be applied for better SNRs as illustrated in of the overall signal. When only noise is present in the total input signal the difference between the NoiseTracker II setting while others prefer being total input signal and the estimated noise will be able to hear more of the environmental noise. small, as will the SNR estimate. Conversely, the Preferences regarding degree of noise reduction SNR estimate will be greater when the total input may also differ from situation to situation, which is signal consists of both speech and noise. the rationale for the different NoiseTracker Depending on the estimated SNR and the user- settings in many of the environmental programs. configurable NoiseTracker II level setting in the For example, the “Traffic” program is set to apply Aventa software, the gain may be reduced as a strong level of noise reduction, as it is assumed illustrated in Figure 4. The gain reduction functions that maximum listening comfort would be desired are mathematically derived using Wiener optimal in this situation. All degrees of NoiseTracker II can filter theory. When noise and speech are present be applied without affecting speech intelligibility simultaneously, the most recent noise estimate is limitations of modulation-based noise reduction systems. Because of the accuracy with which is identifies speech and noise, it excels in its ability to reduce unwanted noise from the incoming signal without affecting audibility of speech or sound quality. With NoiseTracker II, the ReSound hearing instrument wearer can hear desired sounds while noise is kept at comfortable levels to allow a Figure 4. The NoiseTracker gain reduction function is based on the signal-to-noise ratio Bentler, RA (2006). Digital Noise Reduction: an Overview. Trends in Amplification, 10(2), 67- The time constants of the NoiseTracker II system are crucial for its performance. This system Bentler RA, Wu Y, Kettel J, Hurtig R (2008). Digital noise reduction: Outcomes from laboratory and field studies. International Journal of Audiology, 47(8), 447-460. Boll SF (1979). Suppression of acoustic noise in speech using spectral subtraction. IEEE Trans Acoust Speech Sig Proc 27, 113-120. Kochkin S (2000). “Why my hearing aids are in the drawer”: The consumers’ perspective. The
Hearing Journal, 53(2), 34-42.
Kochkin S (2005). MarkeTrak VII: Hearing Loss Population Tops 31 Million People. Hearing Review, 12, (7), 16-29. Kricos P (2006). Audiologic management of older adults with hearing loss and compromised cognitive/psychoacoustic auditory processing capabilities. Trends in Amplification, 10(1), 1-27. Mueller HG, Weber J, Hornsby BWY (2006). The effects of Digital Noise Reduction on the Acceptance of Background Noise. Trends in Amplification, 10(2), 83-93. Nabelek AK, Freyaldenhoven MC, Tampas JW, Burchfield SB, Muenchen RA (2006). Acceptable Noise Level as a predictor of hearing aid use. Journal of the American Academy of Audiology, 17,626-639. Nabelek AK, Tucker FM, Letowski TR (1991). Toleration of background noises: Relationship with patterns of hearing aid use by elderly persons. Journal of Speech and Hearing Research, 34, 679-685. Tun, P., O’Kane, G., Wingfield, A., (2002) Distraction by competing speech in young and older adult listeners. Psychology and Aging, 17, (453-467). Wingfield A, Tun P, McCoy S (2005) Hearing loss in older adulthood, what it is and how it interacts with cognitive performance. American Psychological Society, 14(3), 144-148.

Source: http://www.resound.com.br/~/media/DownloadLibrary/ReSound/White%20papers/noise-tracker-II-white-paper.pdf