Although cochlear implant recipients typically understand speech well in quiet situations, it is well known that many cochlear implant users continue to struggle in noisy situations. Wolfe and colleagues (1,2) measured sentence recognition in quiet and in noise for a group of adults with normal hearing and a group of adults with cochlear implants. Both groups achieved ceiling-level performance (e.g., near 100% correct) in quiet, but at a +5 dB signal-to-noise ratio (SNR), the mean performance of the adults with normal hearing was approximately 95% correct, whereas the mean performance of the cochlear implant recipients decreased to almost 20% correct.
Cochlear implant manufacturers have introduced a number of technologies designed to improve speech recognition in the presence of competing noise. Previous research has shown that these advances, which have primarily consisted of directional microphone technologies and input signal processing that aims to reduce background noise while enhancing speech, possess the potential to improve speech recognition in noise (3–6). For instance, Cochlear Limited introduced SmartSound processing in the Freedom sound processor in 2005. SmartSound processing featured several types of input processing designed to improve performance in challenging listening situations including Adaptive Dynamic Range Optimization (ADRO; which was previously available in the Sprint body-worn processor), Autosensitivity Control (ASC; which was previously available in both the Sprint and Esprit 3G sound processors), Whisper (which was previously available in the Esprit 3G sound processor), and directional microphone (e.g., beamforming/Beam) technologies. The aforementioned input processing schemes may be used in conjunction with one another or in isolation when used with the Freedom sound processor. However, these input processing schemes could only be used in isolation with the Sprint and Esprit 3G sound processors. It should also be noted that Whisper is not intended for use in the presence of competing noise, and because speech recognition in noise is the focus of the current study, Whisper will not be discussed in detail in this paper.
With the introduction of the Nucleus 5 sound processor, Cochlear Limited introduced SmartSound2 input processing. SmartSound2 also included ADRO, ASC, Whisper, and beamforming technology, but because of improvements in the microphone technology of the Nucleus 5 sound relative its predecessor, the Nucleus Freedom, SmartSound2 allowed the recipient to select from two different beamforming technologies known as Beam (which was formerly available in the Nucleus Freedom processor) and zoom. The omni-directional microphones of the Nucleus 5 sound processor were calibrated by the sound processor manufacturer to ensure that they were matched in sensitivity and phase (7). This calibration process, which is also conducted with the Nucleus 6 sound processor, allowed for a more precise and effective implementation of Beam adaptive directionality than what was obtained with the Freedom sound processor. Finally, the Nucleus 6 processor introduced SmartSound iQ input processing, which includes all of the input processing technologies available in SmartSound2 along with two input processing technologies referred to as SCAN and SNR-NR. Each of the input processing technologies mentioned above will be briefly described in the proceeding paragraphs.
Autosensitivity Control (ASC)
ASC is an input processing scheme that automatically adjusts the sensitivity of the sound processor microphone in an attempt to improve comfort and speech recognition in noise. ASC essentially functions as a slow-acting compressor that reduces the gain applied to the microphone signal when the noise level exceeds the breakpoint (i.e., activation threshold which is 57 dB SPL at default settings). The reduction in the gain applied to signal from the microphone shifts the instantaneous input dynamic range (IIDR) of the sound upward to prevent infinite compression for moderate to high-level speech inputs (i.e., prevent the reduction of high-level speech sounds from compression limiting with a high compression ratio from the automatic gain control [AGC] processing). By shifting the IIDR upward, the automatic reduction in microphone sensitivity also prevents lower level inputs from being processed resulting in a reduction in lower level background noise. See Figure 1 for a visual description of how the changes as a function of the sensitivity setting. Wolfe et al. (6) and Gifford and Revit (3) have shown that ASC can provide a significant improvement in speech recognition in noise for adult recipients. Additionally, Wolfe et al. (8) and Gifford et al. (9) reported that the use of ASC improves speech recognition in noise for pediatric recipients.
Adaptive Dynamic Range Optimization (ADRO)
ADRO is a form of input processing that uses slow-acting gain control to continuously adjust the gains across each channel of the sound processor in an attempt to deliver the input signal to the optimal location within the recipient’s electrical dynamic range. ADRO automatically decreases the gain in channels with high outputs to maximize comfort and also to potentially improve speech recognition in noise. ADRO also increases the gain in channels possessing low outputs to enhance audibility of low-level sounds. Research has indicated that the use of ADRO improves the recognition of soft speech (10,11), whereas studies examining the effects of ADRO on speech recognition in noise have been mixed (10,11). Dawson et al. (10) evaluated the benefits of ADRO for a group of 15 children with Nucleus 24 implants and reported that the use of ADRO was associated with a statistically significant mean improvement of 7 percentage points. In contrast, James et al. (11) evaluated the benefits of ADRO for a group of nine adult Nucleus 24 recipients and reported that the use of ADRO typically improved listening comfort in noise without an improvement in sentence recognition in noise.
Researchers have also explored the potential of digital noise reduction/speech enhancement processing to improve speech recognition in noise. One example of this type of pre-processing is Signal-to-Noise Ratio-Noise Reduction (SNR-NR), which is a digital pre-processing algorithm that analyzes the SNR of each channel and automatically reduces the gain of channels comprised primarily of noise. The primary goals of SNR-NR are to improve speech understanding and listening comfort in environments with steady-state noise, such as automobile or industrial noise. Dawson et al. (12) evaluated an early prototype of the SNR-NR algorithm with 13 experienced cochlear implant recipients and found about a 2-dB improvement in the SNR necessary for 50% sentence recognition compared to performance with conventional ACE signal coding. Hersbach et al. (13) also measured word recognition in quiet and sentence recognition in steady state and multi-talker noise with and without SNR-NR for a group of 14 participants unilaterally implanted with a Nucleus cochlear implant. Hersbach and colleagues also measured whether the use of SNR-NR affects music appreciation as well as the potential subjective benefits and limitations of SNR-NR use in real-world environments. Hersbach et al. reported that there were no differences between the SNR-NR enabled and disabled conditions for word recognition in quiet, sentence recognition in multi-talker noise, music appreciation, and subjective preference in quiet environments. However, use of SNR-NR resulted in a statistically significant improvement in sentence recognition in noise (1.3 dB in the speech recognition threshold in noise), and furthermore, the participants preferred to have SNR-NR enabled for noisy situations. In another similar study, Mauger et al. (14) evaluated a prototype of the SNR-NR algorithm and found that SNR-NR provided an average 7 and 27 percentage point improvement in sentence recognition in the presence of babble and steady state, respectively, relative to performance with conventional ACE signal coding. The aforementioned improvements are quite impressive for a single-microphone noise reduction technology. The SNR-NR feature is only available in the Nucleus 6 sound processor.
Directional Microphone Technology (Beamforming)
Numerous researchers have demonstrated that the use of directional microphone technology can offer a substantial improvement in speech recognition in noise for CI users (3–5). Specifically, Spriet et al. (4) evaluated speech recognition threshold (SRT) in noise (50% correct performance) of five Nucleus cochlear implant recipients using the Freedom sound processor, which was the first implant sound processor to implement beamforming technology. Relative to the standard microphone mode (which is slightly directional), Spriet and colleagues reported that the use of beamforming resulted in a mean improvement in the SRT in noise of 15.9 dB when speech babble was presented at 65 dB SPL from one loudspeaker located directly to the side of the implanted ear of the subjects (±90 degrees) and a mean improvement in the SRT in noise of 11.6 dB when speech babble was presented at 65 dB SPL from loudspeakers positioned at 90, 180, and 270 degrees azimuth.
Likewise, Gifford and Revit (3) also showed significant improvement in SRT in noise with the use of beamforming technology in 16 Nucleus recipients using Freedom sound processors. They measured the SRT in noise with the competing “restaurant” noise signal (i.e., a noise signal that was recorded from a typical restaurant environment) was presented from seven loudspeakers positioned 45 degrees apart from 45 to 315 degrees azimuth. Gifford and Revit found a 3.6-dB improvement in the SRT in noise with beamforming enabled relative to the standard microphone condition.
The Nucleus 5 sound processor was the first cochlear implant sound processor to feature two omni-directional microphones that were calibrated to be matched in sensitivity and phase to allow for precise implementation of a directional response via digital subtraction of the electrical outputs of each microphone. This approach allowed for both an accurate execution of both fixed directional polar plot patterns as well as adaptive directionality in which null of the polar plot automatically moved to coincide with direction in the rear hemisphere at which the most intense noise source arrives. SmartSound2 processing in the Nucleus 5 sound processor allowed for the use of three microphone modes, the “standard” mode, which was slightly directional, as well as two beamforming technologies, Beam and zoom.
Beam is an input processing technology that utilizes the matched pair of omni-directional microphones to achieve a directional pattern with a null (or nulls) that automatically varies to be positioned toward the direction at which the most intense noise is arriving. Beam technology incorporates a sophisticated algorithm to identify speech and noise signals and offers attenuation to the latter. Research has conclusively shown that the use of Beam may provide substantial improvement speech recognition in noise (4).
Zoom is also an input processing technology that incorporates the matched pair of omni-directional microphones to achieve beamforming. However, Zoom provides a fixed, highly directional response. Specifically, Zoom creates a fixed, hyper-cadioid polar plot pattern with a maximum null at 120 and 240 degrees azimuth. Wolfe et al. (5) evaluated the benefit of zoom on the speech recognition in noise of 35 unilaterally implanted Nucleus recipients using the Nucleus 5 sound processor. The SRT in noise that was obtained with zoom and ASC + ADRO was compared to the SRT in noise obtained with the standard microphone mode (which is slightly directional) and ASC + ADRO. Speech-weighted noise was presented at 65 and 75 dB SPL directly to the side of the implanted ear (i.e., 90 and 270 degrees azimuth for recipients implanted in the right and left ear, respectively). There was a significant mean improvement of approximately 6 dB with use of zoom over the standard microphone mode with each of the 35 subjects showing the improvement. Zoom and Beam pre-processing are both available in Nucleus 5 and Nucleus 6 sound processors.
Potts and Kolb (15) evaluated sentence recognition in noise of 32 adult Nucleus implant recipients using the Nucleus 5 (CP810) sound processor with various combinations of the input processing schemes ASC, ADRO, Beam, and zoom (e.g., ASC + ADRO, ASC + ADRO + Beam, ASC + ADRO + zoom, Beam alone, zoom alone, etc.). Specifically, HINT sentence recognition was assessed in the presence of restaurant noise presented at 70 dB SPL from eight loudspeakers surrounding the participant in the horizontal plane. Use of Beam and zoom resulted in significantly better sentence recognition in noise when compared to performance obtained in the standard microphone condition. The best performance was generally obtained when Beam and zoom were used with ASC. When used in conjunction with ASC and ADRO, there was no difference in performance obtained with Beam and zoom.
Use of Input Processing to Improve Speech Recognition in Noise in the Real World
Historically, cochlear implant recipients had to manually switch to an alternative program or memory within their sound processor to access pre-processing designed to improve performance in noise. In the authors’ clinical experiences, some recipients may be unable or unwilling to accurately and reliably switch the program of their sound processor. There is a paucity of published reports describing how frequently and effectively cochlear implant users switch to alternative programs to improve performance in challenging situations. Banerjee (16) did study the frequency at which hearing aid users changed the program/memory and/or the volume control setting of their hearing aids. He found that a group of nine hearing aid users remained in the default setting about 85% of the time. When the users did stray from the default settings, about two-thirds of the time they adjusted the volume control setting and only chose to adjust the program about one-third of the time.
In a study of the effectiveness of microphone technologies and signal processing in cochlear implant technology, Gifford and Revit (3) queried a group of Nucleus Freedom users on their tendency to switch from the default program. It was reported that only 2 of 20 subjects claimed to switch to a program with the beamforming directional technology to improve speech understanding in noise. As a result, for many recipients, it may be desirable for the sound processor to analyze the acoustics of the environment and automatically enable microphone directionality and pre-processing that will optimize performance for a given environment.
Acoustic Scene Classification and Adaptive Processing Within the Nucleus 6 Sound Processor
Many modern hearing aids attempt to automatically classify the environment into one of a number of pre-determined acoustic environments (17). In essence, the acoustic characteristics of the current environment is compared to an array of templates to classify the environment as speech in quiet, speech in noise, music, noise only, etc. Noisy environments may be further categorized into specific types of noise such as traffic noise, wind noise, speech noise, etc. This classification is based upon several factors including the intensity, spectral and temporal characteristics of the environment as well as modulation depth and rate of modulation. Once the environment is classified, the hearing aid typically selects features and gain settings that are expected to optimize performance in the presence of the current acoustical conditions.
The Nucleus 6 sound processor possesses an acoustic scene classifier referred to as SCAN (7). SCAN is the first acoustic scene analyzer to be incorporated into a commercially available cochlear implant sound processor. The SCAN system is implemented across three stages. In the first stage, known as feature extraction, the audio signal is analyzed and captured at the processor microphone and multiple, relevant acoustic features are extracted, such as spectral shape, overall level, modulation rate and depth, and tonal information. In the second stage, environment classification, the extracted acoustic cues are compared to environmental templates to determine the current listening environment from a fixed set of acoustic scenes including quiet, speech in quiet, speech in noise, noise, wind, and music. The algorithm that is employed is the best match between the user’s environmental acoustic conditions and the stored classified environments, which were designed using a vast compilation of acoustic data compiled during the development of the pre-processing schemes used in Nucleus sound processors. In the third stage, program selection, the sound processor automatically enables the program/pre-processing/directional microphone response designed to optimize performance for the given environment. For example, if speech in noise was detected in the second stage, the Nucleus 6 sound processor would automatically enable the directional beamforming input processing that would optimize performance in noise (i.e., Beam). The SCAN, SNR-NR,and Wind Noise Reduction (WNR) algorithms are only available in the Nucleus 6 sound processor. WNR is an automatically enabled pre-processing scheme that switches to a more omni-directional microphone mode and reduces output level, particularly in low-frequency channels, when wind noise is detected. The WNR feature was not evaluated at every site in this multi-center study, so it will not be discussed further. It should be noted that similar to the Nucleus 5 sound processor, the Nucleus 6 sound processor possesses two calibrated omni-directional microphones that are matched in phase and sensitivity. Digital subtraction of the electrical outputs of these microphones is used to allow for the implementation of the beamforming input processing schemes, Beam and zoom.
Default Input Processing of Nucleus 5 and Nucleus 6 Sound Processors
The default program of the Nucleus 5 sound processor includes ASC + ADRO along with the standard, slightly directional microphone mode, which from this point forward will be referred to as the “standard” microphone mode. Nucleus 5 users must manually switch to an alternative program to access Beam or zoom beamforming technology. For the default SmartSound iQ input processing of the Nucleus 6 sound processor, Cochlear Limited has incorporated the use of ASC + ADRO, SNR-NR, and SCAN. When Nucleus 6 users are fitted with these default input processing technologies, the appropriate beamforming technology (i.e., Beam for speech in noise or zoom in noise) will be automatically selected in environments containing noise.
The objectives of the current multi-center study were:
- To compare speech recognition in noise with use of the default program (ASC + ADRO with the standard microphone mode) of the Nucleus 5 sound processor versus performance obtained with use of the default settings of the Nucleus 6 sound processor (ASC + ADRO, SNR-NR, and SCAN). It should be noted that use of the Nucleus 5 program with ASC + ADRO will heretofore be referred to as the Nucleus 5 default program, whereas use of the Nucleus 6 sound processor with ASC + ADRO, SCAN, and SNR-NR will heretofore be referred to as the default program of the Nucleus 6 sound processor.
- To compare speech recognition in noise obtained with the Nucleus 6 sound processor in the default program versus the Nucleus 6 sound processor with input processing set to the same default settings of the Nucleus 5 sound processor.
- To evaluate the potential benefit of the SNR-NR noise reduction processing in the Nucleus 6 sound processor.
SUBJECTS AND METHODS
A within-subject, repeated-measures design was used in this multi-center study to accommodate the variability in demographics and outcomes inherent in the population of persons with hearing loss and cochlear implants. Ninety-three subjects were tested at five different sites: 1) Cochlear Americas (Centennial, CO), 2) Cochlear Ltd. (Sydney, Australia), 3) Dallas Ear Institute, 4) Hearts for Hearing, and 5) Houston Ear Research Foundation. Because it is impossible to completely conceal the identity of the Nucleus 5 and Nucleus 6 sound processors, the examiners and subjects were not blinded to the independent variables (e.g., sound processor). This research study was approved by the Western Institutional Review Board (WIRB) and the Royal Prince Alfred Hospital, NSW. The study protocol, along with potential risks and benefits, were thoroughly explained to each subject, and the examiners obtained informed consent before the subjects’ participation.
Ninety-three recipients of the Nucleus Freedom (CI24RE[CA]), Nucleus CI512, and Nucleus 422 cochlear implant systems were included.
Each subject met the following inclusion/exclusion criteria:
- At least 8 years of age at the commencement of the study,
- Implanted with a Nucleus Freedom CI24RE, Nucleus CI512, or Nucleus 422 cochlear implant system and using the Nucleus 5 sound processor,
- Users of the Advanced Combination Encoder (ACE) signal coding strategy and ASC + ADRO pre-processing,
- At least 3 months of experience with the cochlear implant,
- Native speaker of English, the language used to assess speech perception performance, and
- After completion of subject consent and presentation of study protocol, subjects were required to exhibit a willingness to participate and to comply with all requirements of the protocol.
- Unrealistic expectations regarding the potential benefits, risks, and limitations associated with the study procedures and/or equipment, and
- Inability to complete all test sessions.
Of the 93 subjects who participated, 81 were adults and 12 were children. For the purposes of this study, pediatric subjects were defined as those between the ages of 8 and 21 years. The mean age of the subjects was 52 years and 10 months (SD = 22 yr) with a range from 8 to 91 years. The average age of the 12 children was 12 years (SD = 4 yr), and the average age of adults was 58 years (SD = 16 yr).
All procedures were completed within one session for each subject. At the initial study visit, the study protocol was discussed with each subject, informed consent was completed, and a hearing history examination was completed. The subjects’ default program of their Nucleus 5 sound processor was then converted to the Nucleus 6 sound processor using the Nucleus Custom Sound 4.0 programming software. The stimulation levels and additional program parameters (e.g., maxima, frequency allocation, instantaneous input dynamic range, etc.) were identical between the programs loaded onto the Nucleus 5 and Nucleus 6 sound processors. Also, the examiners ensured that the volume and sensitivity controls were set to the same position for the two sound processors. The following listening programs were created on the Nucleus 6 Sound Processor for testing:
- ASC + ADRO: standard microphone directionality
- ASC + ADRO + SNR-NR: standard microphone directionality
- ASC + ADRO + SNR-NR + SCAN: SCAN enabled—automatic microphone directionality (Default Nucleus 6 program)
It should be noted that all subjects were tested with both the Nucleus 5 and Nucleus 6 sound processors immediately after they were fitted with the Nucleus 6 sound processor.
All assessments were conducted in double-walled audiometric test booths. Unilaterally implanted subjects were evaluated with their opposite ear occluded with a foam ear plug to allow for an assessment of their performance with the cochlear implant apart for any acoustic contribution from the contralateral ear. Bilaterally implanted subjects were asked to select the ear that they perceived to be their strongest performing ear, and they were asked to remove their cochlear implant from their opposite ear, which was occluded during assessment. Sentence recognition in noise was evaluated with the AzBio sentences (18–20). Speech-weighted noise served as the competing noise signal. Sentences were presented from a loudspeaker positioned 1 m directly in front of the subjects. The noise signal was presented from a loudspeaker located 1 m directly to the side of the subject’s implanted ear (90 and 270 degrees azimuth for right and left ear implanted subjects, respectively).
Each subject’s performance was initially evaluated with the Nucleus 6 sound processor with ASC, ADRO, and SNR-NR enabled. For the initial assessment, the examiner determined the SNR that resulted in the subject’s performance falling within 40 to 60% correct. This assessment commenced at a +10 dB SNR (sentences presented at 60 dBA). The examiner calculated the participant’s score after five sentences were presented. If the score fell between 40 and 60%, then the examiner presented an additional five sentences to confirm that the total score for the 10 sentences fell between 40 and 60%. The level of the competing noise signal was varied in 2-dB steps until an SNR resulting in performance between 40 and 60% correct was obtained across 10 successive sentences. This SNR served as the baseline SNR at which the remaining test conditions were assessed (heretofore referred to as the “established SNR”). Sentence recognition was evaluated for one list of 20 AzBio sentences at the established SNR while the participants used the Nucleus 6 sound processor with ASC, ADRO, and SNR-NR enabled.
The order in which the remaining test conditions were completed was counterbalanced in an attempt to prevent an order effect from confounding study results. For each test condition, sentence recognition was evaluated for 20 AzBio sentences presented at the “established SNR” (sentences presented at 60 dBA). The test conditions were as follows:
- Nucleus 5 with the default program
- Nucleus 6 with ASC + ADRO and standard microphone mode (SNR-NR disabled)
- Nucleus 6 with ASC + ADRO and SNR-NR enabled with standard microphone mode
- Nucleus 6 with default program
Average percent correct scores on AzBio Sentences in each condition are shown in Figure 2. The group data were analyzed using a within-subjects repeated-measures analysis of variance with one independent variable (test condition). The analysis revealed a significant main effect of test condition, F (1, 371) = 74.3, p < 0.00001. Post hoc comparisons were conducted to more closely examine the main effect of test condition using a Tukey-Kramer multiple comparisons test. According to this analysis, average scores in the Nucleus 5 default program and Nucleus 6 with ASC + ADRO were not significantly different, but both of these conditions yielded significantly poorer average performance (p < 0.05) than the two remaining Nucleus 6 conditions (i.e., Nucleus 6 ASC + ADRO + SNR-NR and Nucleus 6 default program [SCAN + ASC + ADRO + SNR-NR]). Additionally, the use of the Nucleus 6 condition with the default program resulted in significantly better (p < 0.05) average speech recognition performance than the Nucleus 6 condition with only ASC + ADRO + SNR-NR.
To more closely examine individual data in the post hoc comparisons, scatter plots are provided in Figures 3, 4, and 5. In these three figures, individual scores from the three Nucleus 6 conditions are shown as a function of scores in the Nucleus 5 Everyday condition. The dotted lines represent the upper and lower 95% confidence intervals predicted by the binomial distribution for one list of the AzBio sentences (20,21). As a result, figures that show a large proportion of individual data points outside of the dotted lines represent performance differences, for 95 of 100 cases, between the Nucleus 5 and Nucleus 6 conditions. As expected from the post hoc analyses explained above, relative to the Nucleus 5 condition, the largest proportion of individuals had higher performance with the Nucleus 6 when SNR-NR was activated (56%) and particularly when SNR + NR was activated with SCAN (88%).
Because of the wide age range of participants in this study, two additional scatter plots are provided to examine potential effects of age on measured benefit in the Nucleus 6 ASC + ADRO + SNR-NR and Nucleus 6 default program (SCAN + ASC + ADRO + SNR-NR). More specifically, for Figure 6, difference scores were calculated between the Nucleus 6 with ASC + ADRO condition and the Nucleus 6 with ASC + ADRO + SNR-NR. The difference scores, which predict benefit from adding the SNR-NR, were then plotted as a function of participant age in years. Similarly, for Figure 7, difference scores were calculated between the Nucleus 6 with ASC + ADRO condition and the Nucleus 6 default program (SCAN + ASC + ADRO + SNR-NR). These difference scores, which predict the benefit of adding SNR-NR as well as SCAN, were then plotted as a function of participant age. Given the small coefficient of determination (i.e., R2 = 0.027) and the flat trendline shown in Figure 6, there does not seem to be a noteworthy relationship between age and benefit from adding SNR-NR. Likewise, the small coefficient of determination (i.e., R2 = 0.0044) and the flat trendline shown in Figure 7 suggests no relationship between age and benefit from adding SNR-NR as well as SCAN.
The results of this study suggest that the Nucleus 6 sound processor possesses the potential to improve Nucleus users’ ability to understand speech in the presence of noise. When compared to the Nucleus 5 in the default program, the mean sentence recognition score was 27 percentage points higher with the use of the Nucleus 6 in the default program (see Fig. 1). When comparing performance across the different test conditions of this study, it is apparent that the majority of the difference in performance obtained between the use of the default program of the Nucleus 5 and the default for Nucleus 6 sound processors is likely attributable to the automatic activation of the beamforming directional mode with use of Nucleus 6 and SCAN and also to the inclusion of SNR-NR in the Nucleus 6 sound processor. This is not a trivial finding as it represents the first published evidence from a multi-center study with a large number of subjects showing the potential benefit of a cochlear implant sound processor that automatically switches to a more aggressive directional mode when speech is detected in the presence of background noise. This finding is significant because previous research (3) suggests that cochlear implant recipients typically do not manually activate directional programs in their processors in an attempt to improve performance in noisy, real-world conditions. Automatic activation allows a recipient to have optimal access to incidental sounds throughout the environment in quiet situations while also allowing for the potential of better performance in noise with automatic activation of the directional mode.
It is also important to note the significant improvement in sentence recognition in noise observed with the use of the SNR-NR noise cancellation/speech recognition pre-processing feature. On average, the participants scored 9 percentage points higher with the Nucleus 6 sound processor when SNR-NR was enabled along with ASC + ADRO compared to performance with ASC + ADRO alone. This improvement compares favorably to evaluations of single-microphone signal processing strategies designed to improve performance in noise for hearing aid users (22). Research has typically shown that digital noise reduction in hearing aids improves comfort in noise but results in no improvement in speech recognition in noise (22). The improvement observed with use of SNR-NR in noise is similar to what has been reported in other studies that have evaluated the potential benefit of noise reduction strategies for cochlear implant users (23). Collectively, these studies suggest that cochlear implant users may be more likely than hearing aid users to experience improvement in speech recognition in noise with the use of digital noise reduction algorithms. It should be noted that when only ASC + ADRO were enabled in each processor, performance obtained with the Nucleus 5 and Nucleus 6 sound processors was similar. This finding indicates that the adaptive signal processing strategies present in the Nucleus 6 sound processors (e.g., SNR-NR and SCAN) were primarily responsible for the improvements observed with the use of the Nucleus 6 sound processor, and as shown in Figures 6 and 7, age does not seem to impact benefit from SNR-NR and SCAN.
Benefit from the use of SCAN and the automatic activation of beamforming observed in this study may be greater than what a recipient would experience in realistic settings. This study used a single source to present the noise signal, which was directed toward the side of the implanted ear, a position that falls within the region of attenuation for an adaptive directional microphone. Furthermore, testing was conducted in audiometric test booths. It is well known that the benefit from directional microphones diminishes when used in environments characterized by diffuse noise and high levels of reverberation (24,25). Performance in noise with the use of the Nucleus 5 would most likely improve with manual activation of beamforming technology. However, as noted earlier, many recipients do not manually activate beamforming during real-world use (3). On the other hand, the relative benefit of SCAN with the automatic activation of beamforming may have been greater if the assessment within this study had been completed at a higher competing noise level (e.g., 70–75 dB SPL). Most adaptive noise technologies become more aggressive at higher competing noise levels, so it is possible that the noise attenuation provided by SCAN and SNR-NR would have been greater if tested at higher noise levels which are routinely encountered in the real world.
Finally, additional research is needed to evaluate the efficacy of SCAN and SNR-NR during real-world use. Previous research with hearing aids has clearly shown that directional benefit observed in laboratory settings does not always translate to subjective directional benefit in the real world (26,27). Indeed, a field trial of SCAN would further elucidate the potential benefit it can provide for cochlear implant users and also evaluate whether there may be deleterious effects associated with use of SCAN in the real world (i.e., unwanted attenuation of important signals). Additionally, a field trial is needed to demonstrate the potential benefits and limitations of SNR-NR.
In particular, additional assessment of the potential benefits and limitations of SCAN and SNR-NR for children is needed to determine whether these technologies improve real-world performance and, more importantly, to ensure that the use of these technologies is not associated with any detriment in hearing performance (e.g., inability to hear important sounds, such as incidental speech arriving from behind the child) and/or speech, language, and auditory development. Furthermore, it should be noted that the SCAN feature is disabled when a remote microphone radio receiver is coupled to the Nucleus 6 sound processor. In many situations, it is likely that the use of remote microphone technology may provide better hearing performance than the SCAN system (e.g., listening to a teacher lecture in a noisy, reverberant classroom). Further research is needed to better understand the ideal application of SCAN, SNR-NR, and remote microphone technology in the pediatric population.
Given the multi-site nature of this investigation, there were several limitations to the present study. First, test environments varied across sites; however, all testing was conducted in a sound-treated booth. Second, multiple investigators conducted the assessments; however, limited variability was expected given the strict protocol followed by the examiners. Third, participants were not given a period of acclimatization with the new processor; therefore, speech recognition scores reported in the present study may underestimate scores after a period of use. Fourth, pediatric participants were tested using AzBio sentences, which were created for use with adults. A pediatric version of the AzBio sentences (the Baby Bio) (28) does exist, but it was unavailable at the time data collection was completed in this study. However, as indicated earlier, there was not a statistically significant difference in performance between pediatric and adult participants in this study, so the use of AzBio sentences with children does not seem to have had an overly detrimental effect on their performance. Finally, participants using bilateral sound processors were tested in a unilateral condition. This was done to control for any binaural improvements found with the algorithms, but may have caused a detriment in performance in some individuals who relied on binaural benefits when listening in noise (e.g., binaural summation; squelch).
- Mean speech recognition in noise obtained with the Nucleus 6 in the default program (ASC + ADRO, SNR-NR, and SCAN) was 27 percentage points higher than that obtained with the Nucleus 5 in the default program (ASC + ADRO).
- Use of SNR-NR along with ASC + ADRO in the Nucleus 6 sound processor improved speech recognition in noise by a statistically significant average of 9 percentage points when compared to use of the Nucleus 6 with ASC + ADRO alone.
- The SCAN acoustic scene classifier with automatic activation of optimal SmartSound iQ input processing allows for a substantial improvement in speech recognition in noise.
1. Wolfe J, Morais M, Neumann S, et al. Evaluation of speech recognition with personal FM and classroom audio distribution systems. J Educ Audiol
2013; 19: 65–79.
2. Wolfe J, Morais M, Schafer E, et al. Evaluation of speech recognition of cochlear implant recipients using a personal digital adaptive radio frequency system. J Am Acad Audiol
2013; 24: 714–24.
3. Gifford RH, Revit LJ. Speech perception for cochlear implant recipients in a realistic background noise: Effectiveness of preprocessing strategies and external options for improving sentence recognition in noise. J Am Acad Audiol
2010; 21: 441–51.
4. Spriet A, Van Deun L, Eftaxiadis K, et al. Speech understanding in background noise with the two-microphone adaptive beamformer BEAM in the Nucleus Freedom cochlear implant system. Ear Hear
2007; 28: 62–72.
5. Wolfe J, Parkinson A, Schafer EC, et al. Benefit of a commercially available cochlear implant processor with dual-microphone beamforming: A multi-center study. Otol Neurotol
2012; 33: 553–60.
6. Wolfe J, Schafer E, Heldner B, Mulder H, Ward E, Vincent B. Evaluation of speech recognition in noise
with cochlear implants and Dynamic FM. J Am Acad Audiol
2009; 20: 409–21.
7. Mauger SJ, Warren CD, Knight MR, Goorevich M, Nel E. Clinical evaluation of the Nucleus 6 cochlear implant system. Performance improvements with SmartSound iQ. Int J Audiol
2014; 53: 564–76.
8. Wolfe J, Schafer EC, John AB, Hudson M. The effect of front-end processing on cochlear implant performance of children. Otol Neurotol
2011; 32: 533–8.
9. Gifford RH, Olund AP, Dejong M. Improving speech perception in noise for children with cochlear implants. J Am Acad Audiol
2011; 22: 623–32.
10. Dawson PW, Decker JA, Psarros CE. Optimizing dynamic range in children using the Nucleus cochlear implant. Ear Hear
2004; 25: 230–41.
11. James CJ, Blamey PJ, Martin L, Swanson B, Just Y. Adaptive Dynamic Range Optimization for cochlear implants: A preliminary study. Ear Hear
2002; 23: 49S–58S.
12. Dawson PW, Mauger SJ, Hersbach AA. Clinical evaluation of signal-to-noise ratio-based noise reduction in Nucleus® cochlear implant recipients. Ear Hear
2011; 32: 382–90.
13. Hersbach AA, Arora K, Mauger SJ, Dawson PW. Combining directional microphone and single-channel noise reduction algorithms: A clinical evaluation in difficult listening conditions with cochlear implant users. Ear Hear
2012; 33: e13–e23.
14. Mauger SJ, Dawson PW, Hersbach AA. Perceptually optimized gain function for cochlear implant signal-to-noise ratio based noise reduction. J Acoust Soc Am
2012; 131: 327–36.
15. Potts LG, Kolb KA. Effect of different signal-processing options in speech-in-noise recognition for cochlear implant recipients with the Cochlear CP810 speech processor. J Am Acad Audiol
2014; 25: 367–79.
16. Banerjee S. Hearing aids in the real world: Use of multimemory and volume controls. J Am Acad Audiol
2011; 22: 359–74.
17. Dillon H. Hearing Aids
, 2nd ed. 2012; Boomerang Press, Turramurra, Australia.
18. Gifford RH, Shallop JK, Peterson AM. Speech recognition materials and ceiling effects: Considerations for cochlear implant programs. Audiol Neurotol
2008; 13: 193–205.
19. Schafer EC, Pogue J, Milrany T. Equivalency of the AzBio Sentence Test in noise for listeners with normal-hearing sensitivity or cochlear implants. J Am Acad Audiol
2012; 23: 501–9.
20. Spahr AJ, Dorman MF, Litvak LM, et al. Development and validation of the AzBio sentence lists. Ear Hear
2012; 33: 112–7.
21. Thornton AR, Raffin MJ. Speech-discrimination scores modeled as a binomial variable. J Speech Hear Res
1978; 21: 507–18.
22. Bentler RA. Effectiveness of directional microphones and noise reduction schemes in hearing aids: A systemic review of the evidence. J Am Acad Audiol
2005; 16: 477–88.
23. Koch DB, Quick A, Osberger MJ, Saoji A, Litvak L. Enhanced hearing in noise for cochlear implant recipients: Clinical trial results for a commercially available speech-enhancement strategy. Otol Neurotol
2014; 35: 803–9.
24. Ricketts TA. Impact of noise source configuration in directional hearing aid benefit and performance. Ear Hear
2000; 21: 194–205.
25. Ricketts TA, Dahr S. Aided benefit across directional and omni-directional hearing aid microphones for behind-the-ear hearing aids. J Am Acad Audiol
1999; 10: 180–9.
26. Walden BE, Surr RK, Cord MT. Real-world performance of directional microphone hearing aids. Sem Hear
2005; 26: 70–7.
27. Rickett T, Galster J, Tharpe AM. Directional benefit in simulated classroom environments. Am J Audiol
2007; 16: 130–44.
28. Spahr AJ, Dorman MF, Litvak LM, et al. Development and validation of the Pediatric AzBio sentence lists. Ear Hear
2014; 35: 418–22.