You are on page 1of 8

Siemens SoundSmooting

Skepticism Lab Siemens SoundSmoothing Kate Geisen & Jillian Wendel SLHS 605

licensed under Creative Commons Attribution Share-Alike

Siemens SoundSmooting 1. Technical description:

The name of our feature is SoundSmoothing which is a component of the Siemens speech and noise management system. Although Siemens has named this technology SoundSmoothing a more general name for it is Noise Suppression. The Siemens Centra product line was the first to include the SoundSmoothing technology. It is currently available in all Siemens products with the technology levels of 301, 501, and 701. The goal of SoundSmoothing is to reduce the annoyance of transient background noises as they account for one third of the occurrences of everyday noises. This includes reducing noises such as doors slamming, silverware clanking, and paper rustling (Whitepaper, 2008). SoundSmoothing was designed to work with noise suppression technology Siemens already has is place. The first is digital noise reduction algorithms that work based on modulations. These algorithms typically deal with noises that are continuous in time. Spectral Subtraction, which uses a Wiener Filter, was the second form of digital noise reduction implemented by Siemens (Wiener 1946, Hamacher et al., 2005). A Wiener filter is a filter used to reduce the amount of noise occurring in a signal by comparing the signal to an estimation of the signal without noise. It uses fast-acting noise reduction filters that complement the previously in place modulation based algorithms, reducing stationary broadband noise and reducing undesired noise during gaps in speech. However, although Spectral Subtraction is fast acting, there was potential for it to reduce the modulations in speech as well so the new technology was developed to target transient sounds. SoundSmoothing works using the spectral and temporal properties of the input. A block diagram illustrating the technical process can be seen below. The spectral information allows the algorithm to bring down low or high frequency transients without affecting other frequency regions. The temporal information is calculated with high resolution to keep the reaction times down (Whitepaper, 2008). The envelope features are next analyzed to decide if the input contains speech or not and only makes adjusts to non-speech. Using the ratio of peak level to long-term overall RMS the system makes reductions in the amount of gain reduction, providing more reduction as the amount of transients increases. The noise level must reach a criterion threshold or no adjustments will be made so that things like the ticking of a clock will not activate SoundSmoothing but a dropped plate would (Whitepaper, 2008).

Siemens SoundSmooting

In the software SoundSmoothing has three different settings available, which differ by the amount of gain reduction and detection threshold. The Min setting has the smallest amount of gain reduction and highest level of detection threshold, which is what activates SoundSmoothing. This would be the ideal setting for someone who wants to be able to hear the soft transients like a clock. The Max setting has the largest amount of gain reduction and the lowest detection threshold meaning it would be good for those individuals who are very bothered by things like keys rattling or a door slamming. The Med setting is the default setting that is most often used. A tradeoff is that activating SoundSmoothing can add to the delay of the instrument but this added delay is less than 1 ms. Its activation does not degrade any speech present in the input (Whitepaper, 2008). 2. User benefit: For a hearing aid user SoundSmoothing is there to bring down the transient sounds that were not previously covered by the standard digital noise reduction algorithms or by Wiener filtering. As there are three different settings for this feature it will depend which is selected as to how much benefit a user will find in this feature and how often it will be activated. For those users that are very bothered by the transient environment sounds the max setting should be selected to cut them down as much as possible. Enabling this feature does not have any effect on other listening situations as shown in a study completed at the National Acoustic Laboratories in Australia. This study involved a forced choice task for if six scenarios of either preferring SoundSmoothing on or off. The results shows that for transient sounds SoundSmoothing on was preferred and that with stimuli like speech there was no preference, showing that this feature has no negative effects on speech understanding or quality. Enabling this feature did not lead to any tradeoffs that could be found. However, I imagine that enabling it would lead to an increase in current flow and subsequently more battery drain. 3. Experiment design: A technical check of SoundSmoothing would be to perform electroacoustic measurement recordings from KEMAR by using a slim tube behind the ear hearing aid with a closed tip as described by Chalupper and Branda (2008). Test stimuli included real-world sounds like hammering and dishes and synthetic sounds provided added control over the input signal. The stimuli were presented at recorded at different input levels and durations with the

Siemens SoundSmooting

SoundSmoothing set to off, min, med, or max recording the transient attenuation in decibels at each as well as the difference RMS. To check the effectiveness of SoundSmoothing in an experiment I would provide the subjects in with a pair of hearing aids with two programs that were identical in every aspect except for if SoundSmoothing was on or off. Then I would have them use both throughout a trial period, recording which situations one was preferred over the other, and perhaps use SoundLearning to see which program they spent more time using at the end of the trial. A questionnaire could also be created to give to the subjects asking about certain scenarios but they would need to experience the scenarios in each program within a short enough span of time to accurately select which was more pleasant. An experiment by Keidsar et al. set up speech in six various stimuli that included transient and stationary sounds then presented each with SoundSmoothing on and off to determine which was preferred using a forced choice paired comparison test (2007). The stimuli included hammering, door slamming, paper rustling, cutlery, speech and party noise. This shows which situations users show benefit from having SoundSmoothing enabled and which situations they are indifferent to it. 4. Experiment results: In two separate experiments, the SoundSmooting feature was evaluated (Chalupper & Kramer, 2008; Keidser et al., 2007). A group of hearing aid users ranging in age and hearing aid experience were used to evaluate the transient noise reduction algorithm using several different types of stimuli, including transient sounds, like hammering, door slamming, and paper rustling, as well as stationary sounds, like party noise and speech signals. Statistically significant preferences were found for SoundSmoothing on for hammering, door slamming, and paper rustling. However, there were no preferences seen for the stationary noise. Sound localization and word recognition tests were also administered to see the effect of the feature on (Chalupper & Karmer, 2008). No significant effects were present for sound localization or word recognition, indicating that this feature does not affect localization cues or sound quality of the speech signal. A third study was performed by Chalupper & Branda, which compared the SoundSmoothing feature to a competitor transient noise reduction technique (2008). For this investigation, electroacoustic measurements were evaluated using KEMAR with two hearing aids programmed similarly in frequency response and only differing in their transient noise reduction algorithms. Stimuli were presented at different levels and durations and were composed of real life sounds such as hammering and clattering dishes as well as synthetic sounds that could be easily manipulated. In terms of gain reduction, SoundSmoothing was seen to reduce transient sounds by up to 19 dB in comparison to the competitors 8 dB gain reduction, indicating that the SoundSmoothing feature is more effective in reducing transient sounds (Fig 2) (Chalupper & Branda, 2008). Another comparison was made and demonstrated that SoundSmoothing actually detected and attenuated the peaks of transient signals better than the competitor in the presence of background noise (Fig 5). A major difference seen between the two algorithms was that SoundSmoothing employs an adaptive release time between 10-150ms that can adapt to the transient impulse sounds. The competitor algorithm, on the other hand, employs a fixed release time of 60ms that displayed gain reduction even after the transient impulse had occurred (Chalupper & Branda, 2008). Adaptive release times were shown to be more

Siemens SoundSmooting

effective in managing these transient impulses noises, making the SoundSmoothing feature more effective than the competitor.

5. Saboteur: The SoundSmoothing feature has three different settings that differ in the amount of gain reduction and level of activation thresholds. For the min setting, the level of noise required for the feature to activate (detection threshold) is high (60dB) and when detected, the amount of gain reduction is relatively low (20 dB). For the med settings, the detection threshold is 50 dB and the gain reduction is 30 dB; this is the default setting of the hearing aid. The max setting has a much lower detection threshold at 40 dB and greater gain reduction of 40 dB. This setting is optimal for users who are highly annoyed by transient noises, as it is a very aggressive setting. This feature would not be as effective if the level of settings do not match the users preferences & complaints. Mild annoyance by transients/max setting 1. place in location w/transient noises >40 dB 2. have user rate annoyance level 3. have user rate perceived benefit from feature Severe annoyance by transients/ min setting 1. place in location w/transient noises <60 dB 2. have user rate annoyance level 3. have user rate perceived benefit from feature

Siemens SoundSmooting 4. have user rate sound quality level & perceived audibility Prediction of results: 2. would be low because the user has previously indicated mild annoyance 3. could vary since the maximum setting could be attenuating the transients that mildly annoy them. 4. sound quality should be rated as high because SoundSmoothing shouldnt affect sound quality, however audibility could be rated lower if the activation of the feature is attenuating sounds that werent previously annoying them. Bottom Line: If the incoming signal is not bothersome to the individual, we want the hearing aid to have as little effect as possible. The settings were too aggressive for this individuals preferences.

6 4. have user rate sound quality level & perceived audibility Prediction of results: 2. would be high because the user previously indicated severe annoyance 3. would be high because the minimum setting is not attenuating any of the transient noises because the detection threshold has not been met. 4. sound quality should be rated and high and audibility will remain because the feature is not currently activated.

Bottom Line: The incoming signal contains transient noises that are severely annoying to the user. The feature is not activated, due to being set too conservatively for this individual.

Also, we must remember that noise management systems are not perfect and this pertains to the SoundSmoothing feature as well. By using a variety of transient noises, some bothersome transient noises (such as hammering, dishes clinking, clocks ticking) and some necessary transient noises (smoke alarm, dog barking, diabetic pump beeping), you could potentially result in the feature attenuating alarming sounds that are important in alarming an individual. What may be an annoying transient noise to the algorithm could also be an important indicator to the hearing aid user. 6. Judgment: Initially, I thought this feature could be very beneficial for hearing aid users who are annoyed by loud sounds because comfort, as well as audibility, are the hearing aid fitting goals for an audiologist. At first, I thought of this feature as a standalone solution to all transient noise problems, able to detect and erase any transient sound that could be annoying the user. However, the more I learned about it and its effectiveness, the more I realized that it really is more of a supplementary feature to Siemens digital noise reduction system. It acts as an enhancement to the already successful noise management system and specializes in reducing the gain for transient sounds without affecting sound quality. This algorithms efficacy has been shown, however like all noise management systems, there is no single algorithm that can work in conjunction with the users own mind. While modulation detectors do a great job of determining what is speech and what is noise, they fail in situations when what the algorithm considers noise and what the user considers noise are different. After going through this exercise I still believe this to be a very beneficial feature for hearing aid users that are annoyed by transient noises. I think it can

Siemens SoundSmooting

be very helpful in providing comfort to those who might otherwise reject amplification, and for that reason alone it is valid in my book. I was probably most surprised by how little outside research I could find on this feature. All of the empirical evidence I found was from Siemens themselves, which should always be considered with a factor of biasing. A more subjective evaluation from an outside researcher would have been nice to use as a reference.

Siemens SoundSmooting References Chalupper, J., Branda, E. (2008). Comparison of Transient Noise Reduction Systems. Hearing Review, 15(1), 20-24. Chalupper, J., Kramer, B. (2008). Evaluation of a transient noise reduction algorithm. Siemens Audiologische white paper. Retrieved from http://w1.hearing.siemens.com/ch-fr/01-professional/audiologie/02-resultatsetude/_resources/pdf/sr-soundsmoothing.pdf Hamacher, V., Chalupper, J., Eggers, J., Fischer, E., Kornagel, U., Puder H., Rass U. (2005). Signal Processing in High-End Hearing instruments: State of the Art, Challenges, and Future Trends. Journal on Applied Signal Processing, 18, 2915-2929 Keidser, G., OBrien, A., Latzel, M., Convery E., (2007). Evaluation of a noise-reduction algorithm that targets non-speech transient sounds. Hearing Journal, 60(1), 29-39.

Siemens. SoundSmoothing: A New Algorithm to Reduce the Annoyance of Noise. White paper. Retrieved from: http://hearing.siemens.com/Resources/Literature/_Global/publications/2008%20%20SoundSmoothing.pdf?__blob=publicationFile Wiener, N. (1948). Cybernetics.Cambridge, MIT Press.

You might also like