Evaluation of noise reduction technologies in a contemporary cochlear implant system

Abstract

Advances in cochlear implant technology have resulted in excellent outcomes for most pediatric and adult recipients. Specifically, most adult cochlear implant users can converse over the telephone and perform at ceiling levels (i.e., 100%) on open-set sentence recognition tests used for clinical assessment,1-4 while many children who receive a cochlear implant at an early age develop age-appropriate speech, language, and academic abilities.5-7 Despite these advances, many cochlear implant users continue to experience substantial difficulty with speech recognition in noisy environments. In particular, recent studies have shown that speech understanding decreases by 30 to 60 percentage points when performance in quiet is compared with performance at commonly-encountered signal-to-noise ratios ranging from +4 and +10 dB.8-11 As a result, cochlear implant manufacturers invest considerable resources into the development of technologies designed to improve speech perception in noise. For example, the newly-introduced Cochlear Nucleus 5 cochlear implant system possesses several features that are intended to improve speech understanding in noisy environments, which include: Preprocessing strategies designed to augment the speech signal while reducing the salience of the background noise; Microphone technology intended to improve the signal-to-noise ratio to the listener; and Simple and improved connectivity of the implant processor to personal FM systems. The following three sections provide an overview and evidence to support the use of the aforementioned features through discussions of preliminary experiences from studies conducted with children and adults using the Nucleus 5 cochlear implant system. CHANGES IN SIGNAL PROCESSING TO IMPROVE PERFORMANCE IN NOISE Previous studies have shown the importance of input signal processing on speech understanding in difficult listening environments.10-13 Input signal processing refers to modifications that are made to the input signal prior to transformation of the filter outputs to biphasic electrical pulses. Input signal processing may consist of, but is not limited to, adaptive changes in the gain of the signal from the microphone, compression, and summation and cancellation of the signals from multiple microphone systems to achieve directional-specific processing. Adaptive dynamic range optimization One example of input processing in the Nucleus 5 system is the proprietary algorithm referred to as Adaptive Dynamic Range Optimization (ADRO). ADRO adaptively adjusts the gain across all 22 channels by analyzing three aspects of the input level: the overall input level, the level of the competing noise, and the level of the most intense inputs. Based on this analysis, individual gains are then increased or decreased to adjust the frequency response to enhance access to low-level speech inputs and optimize the overall loudness of the signal.12 Continual modification of channel gains using ADRO provides improved sound quality and speech perception in quiet at soft levels, without decreasing performance in noise.12,14 Input processing is required to present the wide range of acoustic signals of interest into the narrow electrical dynamic range that is typical of cochlear implant recipients. The electrical dynamic range refers to the difference in stimulation (in clinical units, which is directly proportional to the intensity of the electrical stimulation in microamperes) between the recipient's threshold of electrical stimulation versus the amount of electrical stimulation that is considered to be loud but comfortable. When converted to decibels, the typical electrical dynamic range of the cochlear implant user is between 6-20 dB.15,16 As such, the relatively wide range of acoustic inputs (approximately 100 dB SPL) a recipient may need to hear must be fitted into the much narrower electrical dynamic range. Instantaneous input dynamic range Compression technology and adjustments in the gain of input signal are employed to optimize delivery of the desired range of acoustic inputs into the recipient's narrow electrical dynamic range. The Nucleus 5 system has a default instantaneous input dynamic range (IIDR) of 40 dB. The IIDR refers to the range of short-term, instantaneous acoustic intensity fluctuations that are mapped without compression (or other types of attenuation, such as Autosensitivity Control) into the recipient's electrical dynamic range at a given point in time. In a quiet environment, the default setting of 40 dB IIDR results in inputs between 25-65 dB SPL being mapped into the recipient's electrical dynamic range (when the default microphone sensitivity setting is used). All inputs below 25 dB SPL are inaudible to the recipient, while all inputs exceeding 65 dB SPL are subjected to infinite compression. A 40 dB IIDR was chosen to encompass the typical range between the peaks and valleys of ongoing speech. This approach seeks to utilize the entire narrow electrical dynamic range to process the ongoing speech signal. Theoretically, this should optimize speech recognition. This approach, however, is not entirely suitable for processing speech in the presence of a high-level competing noise signal. In conditions that would typically be classified as noisy environments (e.g., restaurants, sporting events, vehicles, auditoriums, etc.), the level of the competing noise signal is typically 65 dB SPL or greater.17 When two people communicate in these types of noisy environments, the talker tends to increase the level of his or her voice so that it exceeds the ongoing competing noise. Consequently, the listener typically benefits from being able to listen at a positive signal-to-noise ratio.17 With a fixed 40 dB IIDR, as described above, both the speech and noise signals would be subjected to infinite compression, and any positive signal-to-noise ratio would be eliminated. As such, the speech signal of interest would be embedded into the ongoing noise, and the cochlear implant recipient would be unable to effectively understand speech in noise. Autosensitivity control Another proprietary form of input processing, referred to as Autosensitivity Control (ASC), is used to adjust the overall gain of the signal from the processor microphone to prevent excessive compression of the speech signal in the presence of moderate- to high-level noise. Specifically, ASC operates to adjust the sensitivity of the processor microphone by measuring the noise floor of the surrounding acoustic environment and reducing the microphone gain when the noise floor reaches a specific intensity level. In contrast to a noise reduction strategy, ASC could alternatively be described as a “speech preservation in noise” strategy. ASC automatically shifts the 40 dB IIDR so that it is centered around the speech signal. In essence, this approach optimizes the delivery of the speech signal into the narrow electrical dynamic range when the recipient is communicating in a noisy environment. See Figure 1 for a visual representation of how the position of the IIDR is modified with changes in the sensitivity setting.Figure 1: Illustration of the effect of microphone sensitivity setting on the position of the IIDR.Wolfe et al. studied the potential of ASC to improve speech understanding in noise for a group of adult cochlear implant users.18 Specifically, the signal-to-noise ratio (in dB) required for 50 percent correct performance on the Bamford-Kowal-Bench Sentences in Noise (BKB-SIN) test was determined for a group of adult cochlear implant recipients. The presentation level of the BKB-SIN sentences was fixed at 75 dBA and two list-pairs were administered to each subject. Data were collected while the subjects used ADRO alone and ASC+ADRO. Figure 2 provides means scores for six adult recipients with bilateral Nucleus cochlear implants. The 3.8 dB improvement in the signal-to-noise ratio provided by ASC+ADRO compared with the ADRO alone condition will offer approximately a 30 percentage point improvement in speech recognition in a difficult listening situation with competing noise.19Figure 2: Mean signal-to-noise ratio (dB) for 50% correct performance on the BKB-SIN test for 6 adults with bilateral cochlear implants.A Students t-test indicated that the difference between the ADRO alone and ASC+ADRO conditions is statistically significant (p<0.05). Since ASC only operates in the presence of moderate- to high-level noise, it will not degrade speech recognition in quiet. This statement is confirmed by the mean score of 90% achieved by listeners using ASC+ADRO on the CNC monosyllabic word recognition test in quiet. Optimization of speech recognition in quiet and noise with use of ASC was also supported by another study by Wolfe et al. with adult Nucleus Freedom users.11 A separate study by Wolfe and colleagues evaluated the use of ASC to improve speech understanding performance in noise for a group of children with Nucleus cochlear implants.20 A similar test set-up as described above was used to evaluate speech recognition in noise for children using ADRO alone and ASC+ADRO. Figure 3 shows mean performance for these two conditions for six children using bilateral Nucleus cochlear implants. The 3.1 dB improvement the signal-to-noise ratio provided in the ASC+ADRO condition compared with ADRO alone will provide up to a 20-25 percentage point improvement in speech recognition in difficult listening situations with competing noise. A Students t-test indicated that the difference between the ADRO alone and ASC+ADRO conditions is statistically significant (p<0.05) for pediatric users as well.19Figure 3: Mean signal-to-noise ratio (dB) for 50% correct performance on the BKB-SIN test for 6 children with bilateral cochlear implants.Figure 4 shows mean word recognition in quiet for the PBK-50 test with a group of children using ASC+ADRO. As seen in this graph, open-set speech recognition in quiet is excellent and does not appear to be adversely affected by the use of ASC. In the Nucleus 5 sound processor, ASC and ADRO are used simultaneously in the Everyday program, which is the default program for use in most regular situations. The collective use of these two pre-processing strategies should optimize listening comfort, access to soft speech, and speech recognition in noise for Nucleus 5 users.Figure 4: Mean speech recognition (percent correct) on the Phonetically Balanced Kindergarten-50 monosyllabic word recognition test administered to six bilaterally implanted children using ASC+ADRO.MICROPHONE TECHNOLOGY FOR ENRICHMENT OF IMPLANT PERFORMANCE IN NOISE The Nucleus 5 sound processor is the first implant sound processor to employ two omni-directional microphones from which the output can be summed to provide directional-specific processing. Specifically, the two-microphone system allows for the implementation of directional algorithms that are utilized to focus on sounds arriving from the front of the recipient and attenuate sounds arriving from the sides and behind the listener. The premise of this strategy is the assumption that the signal of interest arrives from the front of the user and that unwanted, competing noise arrives from the sides and behind the user, as typically occurs during conversations in noisy environments. The two omni-directional microphones of the Nucleus 5 sound processor are matched in phase and sensitivity. The microphone calibration and matching process is performed in-house at Cochlear with each individual sound processor to ensure an optimal frequency response as well as an optimal directional response. There are multiple proprietary directional algorithms available in the Nucleus 5 sound processor. The zoom algorithm provides a fixed supercardioid directional response, which substantially attenuates sounds arriving from behind and to the side of the wearer. The zoom option is most appropriate for moderate- to high-level noise environments in which the competing noise signal is stationary in location. The zoom algorithm is activated in the Noise program of the Nucleus 5 sound processor. (See Figure 5 for a visual representation of the zoom mode in the Nucleus 5 Noise program).Figure 5: In situ polar plot patterns for the Nuclear 5 sound processor in zoom mode (red line) and the freedom sound processor in the noise program.Beam An alternative directional algorithm, referred to as Beam, adaptively positions the null of the directional pattern so that maximal attenuation is provided in the direction of the most intense noise source arriving from behind or to the side of the wearer. Use of Beam is recommended when the interfering noise source changes or is moving throughout the environment. The Beam algorithm was available in the previous generation Cochlear sound processor, the Nucleus Freedom, but improvements in digital signal processing and microphone design in the Nucleus 5 have resulted in more precise positioning of the null to correspond with the direction of the most offensive noise source as well as more precise focus toward signals arriving from in front of the user. The Beam algorithm is activated in the Focus program of the Nucleus 5 sound processor. A multi-center study is currently underway to evaluate the potential benefit of the noise program with zoom in the Nucleus 5 sound processor. At this point in time, six subjects with Nucleus 5 sound processors have been evaluated at the Hearts for Hearing Foundation. Speech recognition in noise was evaluated using the BKB-SIN test, which determines the signal-to-noise ratio required for 50 percent correct performance at each of two different sentence presentation levels, 65 and 75 dB A. Performance was evaluated when subjects used the Noise program (ASC+ADRO) of the Freedom processor and the Noise program (zoom+ASC+ADRO) of the Nucleus 5 sound processor. Mean speech recognition in noise scores for these two conditions are provided in Figure 6. As shown, the use of zoom in the Nucleus 5 sound processor offers approximately 7-8 dB improvement in the signal-to-noise ratio. Performance with the Noise program of the Nucleus 5 sound processor is significantly better than performance with the Noise program of the Freedom processor at each of the two sentence presentation levels (p<0.05).Figure 6: Mean signal-to-noise ratio (dB) for correct performance (%) on the BKB-SIN test for 6 adults with unilateral cochlear implants using the Noise program in the Freedom and Nuclear 5 sound processors.ENHANCING COCHLEAR IMPLANT USERS' ACCESS TO PERSONAL FM SYSTEMS The Nucleus 5 sound processor possesses several unique features which make it very amenable to use with personal FM systems. First, it automatically detects the presence of direct auditory input (DAI) signals, so when an FM receiver is plugged into a specialized FM adapter (see Figure 7), the user automatically hears the FM signal without the need to switch to an FM program or activate the DAI signal. Furthermore, ASC is applied to the signal from the sound processor microphone prior to the mixing of that signal with the FM signal. Then, ASC is applied again to the mixed signal. Theoretically, the first implementation of ASC optimizes the SNR for inputs at the sound processor microphone, while the second application of ASC optimizes the SNR for the combined signal.Figure 7: The Nuclears 5 sound processor with the Phonak ML14i Dynamic FM receiver.An ongoing study is currently being conducted at the Hearts for Hearing Foundation to evaluate Nucleus 5 users' experience with personal FM systems. We will provide a comprehensive report on the results of that study when it is completed, however early results are encouraging. Preliminary data for three bilateral users suggest FM performance in noise (see Figure 8) may be significantly better than what was obtained with the Freedom sound processor and what was observed in previous studies of performance of cochlear implant users with Dynamic FM.Figure 8: Speech recognition in noise (% correct on the HINT obtained for there bilateral recipients using the Nucleus 5 sound processors.CONCLUSIONS Exclusive input signal processing in the Nucleus 5 sound processor allows for significant improvements in speech understanding in noise for both children and adults relative to previous technologies, while maintaining excellent speech recognition in quiet settings. The use of advanced directional microphone algorithms in the Nucleus 5 sound processor also allows for substantial improvement in speech understanding in noise relative to previous technology. The Nucleus 5 sound processor also features design improvements which simplify connectivity to a personal FM system while also improving performance in challenging listening environments. Studies are ongoing to evaluate these features and optimize FM use for Nucleus 5 sound processor recipients. WHAT YOU SAID Last month we asked readers to take our online poll after reading the cover story, “Compound variables continue to cloud AuD extern landscape.” In response to the question, does the AuD externship program need to be revised, 63 respondents answered “yes” and 6 replied “no.”Figure