Assessing the benefit of adaptive null-steering using real-world signals

Abstract

Abstract

This study compared the noise reduction of adaptive null-steering and near-hypercardioid directional hearing-aid algorithms via performance on real-world signals. Using subject-individualized and generic (i.e. similar to current hearing aids), off-line frequency-domain implementations, we processed recordings made through two microphones of a BTE device worn by five subjects. Recording scenarios included homes, offices, cafés, streets, buses, and automobiles. We found practically all (> 95% of recording time) adaptive noise-reduction benefit for generic implementations is below 1.2 dB, and 96% and 92% is below 2 dB for 16-and 32-band individualized implementations, respectively. A 256-band, individualized implementation showed a majority of benefit between 1–4 dB. We found no extended (> 2 s) continuous periods of significant (> 2 dB) benefit for the generic adaptive implementations. The recordings—having many independent and simultaneously active sources, spatially extended sources, significant reverberation, or combinations thereof—indicate an environment comprising few instances of high direct-to-diffuse energy situations. Combined with results from previous field trials, the evidence suggests that such an environment is common and represents a significant limitation on adaptive benefit.

Sumario

Este estudio compara la reducción del ruido en algoritmos de auxiliares auditivos con direccionamento anulado o direccionamiento casi-hipercardioides por medio del rendimiento con señales de mundo real. Usando implementaciones fuera de línea en campos frecuenciales, individualizadas y genéricas (p.ej. similares a las de los auxiliares auditivos usuales), procesamos los registros realizados con dos micrófonos de un instrumento BTE usado por cinco sujetos. Los escenarios de registro incluyeron hogares, oficinas, cafés, calles, autobuses y automóviles. Encontramos prácticamente en todas las implementaciones (> 95% del tiempo de registro) que el beneficio en la reducción de ruido adaptable es inferior a 1.2 dB y que en 96% y 92% es menor a 2 dB para implementaciones individualizadas en bandas de 16 y 32, respectivamente. Una implementación individualizada en una banda de 256, mostró la mayoría de los beneficios entre 1-4 dB. No encontramos períodos continuos extendidos (> 2 s) de beneficio significativo (> 2 dB) con las implementaciones genéricas adaptables. Los registros –que tienen muchas fuentes independientes o simultáneamente activas, extendidas espacialmente, con reverberación significativa o combinaciones de los mismos- indican un ambiente que comprende pocas instancias de situaciones de alta energía de difusión directa. La evidencia, combinada con los resultados de pruebas libres previas, sugiere que este ambiente es común y representa una limitación significativa para el beneficio de adaptación.

Key Words:

• Hearing aid
• Directionality
• Beam-forming
• Null steering

Acknowledgments

A preliminary version of this work was presented at the August, 2008 International Hearing Aid Research Conference in Lake Tahoe, California, USA. We thank two anonymous reviewers for comments that helped improve this work.

Notes

1) In our setup the first room reflection to reach an aid would be from the floor, and would arrive approximately 6.9 ms after the direct wave arrival.

2) In general the convergence time of the adaptive systems used here is on the order of this time constant, i.e. 50 ms. For example, an exponential decay with a 50 ms time constant is a reasonable description of the time-dependence (on a linear time scale) of directional gain in dB in response to a step-change in the incidence angle of an anechoic source from 0 to 180°.

3) Individualized and generic implementations will both sometimes yield non-unity target gain due to limited frequency resolution. Also, this shortcoming is worsened in the generic implementations because they are designed in free space but used on a head.

4) Killion et al (1998) report a 4.7-dB articulation-index weighted directivity index (AIDI) on a manikin, and a hyper-cardioid is known to generate an AIDI of 6 dB in a free-space diffuse field.

5) The maximum possible number of streaks M, of a given duration D s, is given by the formula M = T/(D + 0.0094), where T is the total duration (in seconds) of a given continuous recording. The 0.0094 arises from the fact there are 9.4 ms between frames and there must be one frame of below-threshold benefit between each consecutive streak in the case of the maximum possible number. For our set of recordings, M decreases linearly on a log-log scale from M = 12 660 for D = 1 s, to M = 68 for D = 128 s. M drops more quickly at higher D values, with M = 37 and M = 23 for D = 181 s and D = 256 s, respectively.

6) The field-trial reports show no tests of the aid directionality per se (e.g. via polar patterns). Ricketts and Henry (2002), however, demonstrated improved speech intelligibility in noise over omnidirectional processing using aids from the same manufacturer as those used in Bentler et al (2004). Also, Bentler et al (2006) demonstrated improved speech intelligibility in noise over omnidirectional processing using the aids from Palmer et al (2006). These aids, however, only allowed automatic switching between omnidirectional and directional modes when a directional mode was tested. Thus, directionality may not have been active in all instances where benefit for the adaptive case might obtain.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.