Hearing Aids(noun, plural)
1 (archaic) : electronic devices usually worn by a person for amplifying sound before it reaches the auditory receptor organs.
2 (modern) : electronic devices worn by a person for processing sound to provide audibility and improve the signal-to-noise ratio before it reaches the auditory receptor organs.Introduction:
In Part One of this article, we present our argument for the definition of modern hearing aids to expand from amplifiers to signal-to-noise ratio enhancers. In Part Two, we propose, compare, and contrast the Noise Reduction Index (NRI) and the Average Angular Attenuation (A3), two methods to evaluate the effects of adaptive hearing aid signal processing found in modern hearing aids. In Part Three, we discuss the new Sonic Innovations Innova™ hearing aid, with a NRI of +10 dB SNR and an A3 value of +11 dB.
Part One:
The objective of hearing aid fittings in years past was to improve audibility by manipulation of gain, frequency response, and output. Although those goals had merit in their day, that goal and clinical rationale is now incomplete. The objective of today's hearing aid fittings goes beyond audibility-based prescriptive fittings, because modern hearing aids can and must improve the signal-to-noise ratio (SNR) for conversational speech in the presence of competing background noise. Our new objective serves to satisfy the need put forward by Kochkin (2000) that "the essence of a hearing aid is to improve speech intelligibility in listening environments important to the user."
Support for our definition of modern hearing aids (see Abstract, point 2, above) comes from a recent research report by Galster & Ricketts (2004). Two digital signal processing (DSP) hearing aids were compared: an 8-channel system with slow time constants to a 9-channel system with fast time constants. Hearing aid fittings with the two systems were equated for gain and output (RMS difference Using only multi-channel compression (MCC) the subjects had 1.0 dB SNR better S-I-N with the 9-channel system than the 8-channel system. With noise reduction (NR) activated, the 9-channel system improved S-I-N by an additional 1.3 dB whereas the 8-channel system degraded S-I-N by 0.7 dB. The overall result was a net 3 dB SNR improvement in S-I-N with the 9-channel hearing aid over the 8-channel aid.
Contrary to traditional audiology theory, even though the two hearing aid fittings were equated for audibility in terms of prescriptive fitting, there was a significant difference in S-I-N performance, despite both DSP systems having MCC and NR. We propose this behavioral difference was a result of the differing abilities of the two hearing aids to process speech and noise in real time, to favorably change the SNR, and to improve S-I-N for the listener.
Support for our proposal that the two systems under investigation have different abilities to change the SNR is found in benchtop electroacoustic measurements from Hagerman et al (2002) and Hayes (2001, 2002).
Hagerman et al used a phase-reversal technique (described in Part Two of this article, below) to measure the physical change in the SNR attributable to DSP NR algorithms. They used two signals, ICRA speech and Esperanto speech, against three broadband maskers, ICRA speech noise, ICRA 6-talker babble, and print-shop noise. They evaluated five modern hearing aids that were equated for audibility, including the two systems utilized by Galster and Ricketts. For ICRA speech masked by ICRA noise, Hagerman et al reported a 4.3 dB SNR improvement for the 9-channel DSP NR algorithm vs. a 1.3 dB SNR improvement for the 8-channel DSP NR algorithm.
Hayes measured RMS levels of phonemes and noise between speech segments to determine the physical change in the SNR attributable to DSP NR algorithms. Specifically, he used 20 items from the FAAF test as the signal against the five environmental maskers of jet cabin noise, oscillating fan noise, ocean waves noise, rain on a tin roof noise, and subway noise. He evaluated four modern hearing aids, equated for audibility, including the two systems utilized by Galster & Ricketts. For the FAAF speech against the five maskers, Hayes reported a 4.9 dB SNR improvement for the 9-channel DSP NR algorithm vs. a 0.2 dB SNR improvement for the 8-channel DSP NR algorithm.
Table 1: Three SNR measures showing increased SNR benefit for NR algorithm in the 9-channel DSP hearing aid with fast time constants over the NR algorithm in the 8-channel DSP hearing aid with slow time constants. See text for details.
Note to Table 1: The 9-channel DSP hearing aid used in all three studies was the Sonic Innovations Natura®. The 8-channel DSP hearing aid used in all three studies was from a major European manufacturer.
It is apparent from Table 1 (above) that behavioral benefit for S-I-N for modern hearing aids are associated with benchtop electroacoustic measures of change in SNR, even when the change in SNR is achieved by methods other than conventional directional microphone technology.
Part Two:
To quantify hearing aid performance, we will address two techniques that measure the influence of hearing aid signal processing on auditory signals, including adaptive algorithms found in modern hearing aids.
The first technique is the newly-developed measure of change in SNR at the level of the hearing aid: the Noise Reduction Index (NRI) (Nilsson & Bray, 2003). The NRI is an objective, benchtop measurement based on methodology originally proposed by Licklider (1948) and refined by Hagerman et al (2002). The measurement is conducted by making two broadband recordings of the output of the hearing aid, one using an input of speech in noise (A) and the other with an input of speech in phase-inverted noise (B). The sum of the two output recordings [(A+B)/2] cancels noise, leaving an estimate of the speech present. The difference between the two output recordings [(A-B)/2] cancels speech, leaving an estimate of the noise present. The ratio of the combined signals {[(A+B)/2] / [(A-B)/2]} allows one to calculate the SNR coming out of the hearing aid. This output SNR may be compared to the input SNR and the resulting change in the SNR is the NRI.
While Hagerman et al utilized this methodology to evaluate the effect of NR algorithms in DSP hearing aids, the NRI technique can be used to quantify any signal processing effect in modern hearing aids, including MCC, NR, directionality, or combinations of signal processing features.
In our examination of six premium DSP hearing aids that were equated for prescriptive audibility, we found NRI values ranging from -1 to +1 dB SNR for MCC, 0 to +4 dB SNR for NR, and +3 to +6 dB SNR for directionality. Combining these NRI measures with behavioral measures across a group of hearing impaired subjects, we determined that a 2 dB SNR change in the NRI correlates to about 1 dB SNR change in S-I-N (R2 = 0.44, p
The second technique to quantify performance is directional microphone responsiveness, and includes methodologies to calculate directional response charts (polar plots), the directivity index (DI) value at a specific center frequency of a one-third octave band, the unweighted average of DI values from 500 Hz to 5000 Hz, the articulation index (AI) weighed average of the DI values from 200 Hz to 5000 Hz, or the speech intelligibility index (SII) weighed average of the DI values from 160 to 8000 Hz (ANSI S3.35 - 2004).
While the average DI, AI-DI, and SII-DI calculations are not SNR measures, they are considered to be of value in predicting the speech intelligibility performance of hearing aids in noisy situations. For example, "every dB improvement in AI-DI results in an increase of speech understanding (for sentence material) of approximately 10 percentage points" (Dillon, 2001, page 28). For such tests, "the P-I function has a slope of at least 10% per dB of intensity or SNR" (Dillon, 2001, page 392) leading to the very useful generalization that a 1 dB change in AI-DI is predictive of a 1 dB SNR change in S-I-N.
Average DI values are limited to prediction of performance in diffuse, or at least, relatively uniform, noise fields. It is also desirable to quantify the attenuation provided by directional systems when all noise sources arrive from a narrow angular range (e.g. a fairly discrete noise source). With the desire to quantify the average performance of adaptive algorithms in such environments, the directionality measures in Part Three (below) were obtained using the planar directivity index method using articulation index weighting in the standard (ANSI S3.35 - 2004). Since these measures are aimed at predicting performance in discrete rather than uniform noise sources, applying the AI-DI label to such measurements is inappropriate, and they may be more accurately described as "average angular attenuation" (Ricketts, 2005), which we will shorten to AAA or A3.
The A3 values obtained in a planar field comparing fixed vs. adaptive directionality can be used to compare signal processing effects, and may be indicative of greater SNR enhancement for the adaptive algorithm in discrete fields. Furthermore, A3 values that are greater for adaptive directionality over fixed directionality may be predictive of increased benefit for a moving masker, such as a car passing by a hearing aid user.
The two methods of NRI and A3 are similar in that they are sound field measurements (Ghent, 2005; Nilsson et al, 2005) that quantify with electroacoustics the performance of the hearing aid, generating a decibel figure of merit. However, the two methods differ, as described in Table 2 (below).
Table 2: Comparison of two benchtop measures used to quantify signal processing effects. NRI is the Noise Reduction Index, a measure of change in the signal-to-noise ratio. A3 is the Average Angular Attenuation, a measure of directionality.
Part Three:
The new Innova hearing aid from Sonic Innovations provides multiple DSP algorithms designed to electroacoustically change the SNR, thereby facilitating an improvement in S-I-N. Examples of algorithms with these desirable capabilities are D1 Configurable Directionality, D2 DIRECTIONALplus™, and D3 DIRECTIONALfocus™ (Bray, 2005; Nilsson, 2005).
D1 Directionality: The polar patterns available with Configurable Directionality allow the clinician to select the base directivity for a given listening situation. The three available directional options are hypercardioid, supercardioid, and dipole. For example, regarding the hypercardioid pattern, the NRI value is 4.7 dB SNR and the A3 value is 5.7 dB, as shown in Table 3. These NRI and A3 values are achieved using the conventional directional microphone benefits, whereby sounds from behind are de-emphasized, while sounds from the front are emphasized.
Table 3: NRI and A3 measures for D1, D2, and D3, three signal processing modes designed to improve speech understanding in noise and found in the new Innova hearing aids from Sonic Innovations. See text for details.
Notes to Table 3: The noise reduction algorithm is enabled for NRI measures and disabled for A3 measures. The A3 value of 5.7 dB for D1 and D2 modes, obtained with fixed directionality, may also be labeled as AI-DI values.
D2 Directionality: The second level of directionality integrates Adaptive Noise Reduction with Configurable Directionality and the combination is termed DIRECTIONALplus in Natura hearing aids. DIRECTIONALplus is the only DSP platform which shows a significant increase in S-I-N benefit for directionality + NR, as compared to directionality alone, or NR alone (Bray & Nilsson, 2001; Murphy et al, 2002; Nilsson & Bray, 2002). For D2 Directionality in Innova, the A3 value is unchanged. However the NRI value is increased from the D1 value to 7.8 dB SNR (see Table 3, above). The NRI value increases by 3 dB SNR because the NR algorithm is a SNR enhancer, de-emphasizing the steady-state masker regardless of direction, while emphasizing the modulated sentences.
D3 Directionality: The highest level of directionality combines the new DIRECTIONALfocus algorithm with Adaptive Noise Reduction and Configurable Directionality. DIRECTIONALfocus increases directionality through an adaptive algorithm based on the direction of sound, utilizing azimuth-dependent attenuation instead of the traditional adaptive approaches of 'null steering' or 'program switching.' This takes place in each of Innova's 16 channels of compression for both modulated and steady-state sounds.
As shown in Table 3, for D3 Directionality the NRI value is improved to 10.4 dB SNR, using the hypercardioid directionality pattern and the Small Group DIRECTIONALfocus algorithm. The A3 value for D3 Directionality is improved to 11.1 dB, using the dipole directionality pattern and the Single Speaker DIRECTIONALfocus algorithm. Both the NRI and A3 values are improved over the D2 value because the DIRECTIONALfocus algorithm adaptively de-emphasizes sounds from the sides and behind, while emphasizing sounds from the front.
D3 Directionality combines three of Innova's key technologies to create a robust directional system that provides previously unmatched directional performance. Innova's NRI value of +10.4 dB SNR is unprecedented and extraordinary, especially when compared to the NRI measures of between +3 and +6 dB SNR for competitive DSP hearing aids (Nilsson & Bray, 2004). Innova's A3 value of +11.1 dB establishes the hearing aid as unique among adaptive directionality instruments.
Summary:
We propose that the definition of modern hearing aids should be expanded from devices that provide audibility to devices that also favorably change the SNR for the benefit of the listener. Evidence for our proposal is found in the benchtop SNR measures of Hagerman et al (2002) and Hayes (2001, 2002), combined with the behavioral S-I-N measurements of Galster & Ricketts (2004) showing significant differences comparing two DSP hearing aids, despite the aids generating equal amounts of prescriptive amplification.
We have standardized the NRI as a new tool for measuring the SNR change at the level of the hearing aid and have shown this can be a significant predictor of S-I-N performance (Nilsson & Bray, 2003, 2004).
Based on NRI measures with a 0 dB SNR acoustic signal input to the Innova BTE, the hearing aid output is +4.7 dB SNR for D1 Directionality, +7.8 dB SNR for D2 Directionality, and +10.4 dB SNR for D3 Directionality. For Innova's D3 Directionality with DIRECTIONALfocus, the NRI value of +10 dB SNR and the A3 value of +11 dB have set new performance standards for the hearing aid industry.
References:
ANSI S3.35 - 2004: Method of Measurement of Performance Characteristics of Hearing Aids Under Simulated Real-Ear Working Conditions.
Bray (2005). New Digital Aid Designed for Enhanced Directionality. The Hearing Review, 12(1), 44-47.
Bray & Nilsson (2001). Additive SNR Benefits of Signal Processing Features in a Directional DSP Aid. The Hearing Review, 8(12).
Dillon (2001). Hearing Aids. Thieme: New York.
Galster & Ricketts (2004). The Effect of Digital Noise Reduction Time Constants on Speech Recognition in Noise. Research poster at IHCON, Tahoe City, California.
Ghent (2005). A Tutorial on Complex Sound Fields for Audiometric Testing. JAAA, 16, 18-26.
Hagerman, Olofsson & Nästén (2002). Noise reduction measurements in hearing aids. Presentation at IHCON, Tahoe City, California.
Hayes (2001). The Effect of Crossover Frequency on Aided Speech Perception in the Presence of Environmental Sounds. Dissertation for the Department of Communication Sciences and Disorders of the College of Allied Health Sciences, University of Cincinnati.
Hayes (2002). Multichannel Processing of Speech & Environmental Sounds. Presentation at AAA, Philadelphia, Pennsylvania.
Kochkin (2000). MarkeTrak V: "Why my hearing aids are in the drawer": The consumer's perspective. The Hearing Journal, 53(2), 34-42.
Licklider (1948). The influence of interaural phase relationships upon the masking of speech by white noise. JASA, 20, 150-159.
Murphy, Bray, Nilsson & Wellington (2002). Effects of Noise Reduction & Directionality in ITEs. Research poster at AAA, Philadelphia, Pennsylvania.
Nilsson (2005). Sonic Innovations new product - Innova. Interview for Audiology Online, in press.
Nilsson & Bray (2002). Effects of Noise Reduction & Directionality in BTEs. Research poster at AAA, Philadelphia, Pennsylvania.
Nilsson & Bray (2003). Bench Top Index of Noise Reduction Affectivity. Presentation to AAS, Phoenix, Arizona.
Nilsson & Bray (2004). The Noise Reduction Index: Benchtop Estimate of SNR Changes. Research poster at IHCON, Tahoe City, California.
Nilsson, Ghent, Bray & Harris (2005). Development of a Test Environment to Evaluate Performance of Modern Hearing Aid Features. JAAA, 16, 27-41.
Ricketts (2005). Personal communication.
Authors Note: The authors wish to acknowledge and thank Todd Ricketts and Robert Ghent for their valuable comments on an earlier version of this paper.
Editor's Note: Drs. Bray and Nilsson work in the Auditory Research department of Sonic Innovations. Victor Bray is the Vice President for Quality Systems and Michael Nilsson is the Manager for Auditory Research.
