Julie E. Greenberg

Evaluation of an adaptive beamforming method for hearing aids

In this paper evaluations of a two-microphone adaptive beamforming system for hearing aids are presented. The system, based on the constrained adaptive beamformer described by Griffiths and Jim [IEEE Trans. Antennas Propag. AP-30, 27-34 (1982)], adapts to preserve target signals from straight ahead and to minimize jammer signals arriving from other directions. Modifications of the basic Griffiths-Jim algorithm are proposed to alleviate problems of target cancellation and misadjustment that arise in the presence of strong target signals. The evaluations employ both computer simulations and a real-time hardware implementation and are restricted to the case of a single jammer. Performance is measured by the spectrally weighted gain in the target-to-jammer ratio in the steady state. Results show that in environments with relatively little reverberation: (1) the modifications allow good performance even with misaligned arrays and high input target-to-jammer ratios; and (2) performance is better with a broadside array with 7-cm spacing between microphones than with a 26-cm broadside or a 7-cm endfire configuration. Performance degrades in reverberant environments; at the critical distance of a room, improvement with a practical system is limited to a few dB.

INTELLIGIBILITY-WEIGHTED MEASURES OF SPEECH-TO-INTERFERENCE RATIO AND SPEECH SYSTEM PERFORMANCE

This Letter describes decibel measures of speech‐to‐interference ratio and of system gain derived from simple modifications of the articulation and speech transmission indices. The rationale for the use of these measures is presented.

Analysis of speech-based Speech Transmission Index methods with implications for nonlinear operations.

The Speech Transmission Index (STI) is a physical metric that is well correlated with the intelligibility of speech degraded by additive noise and reverberation. The traditional STI uses modulated noise as a probe signal and is valid for assessing degradations that result from linear operations on the speech signal. Researchers have attempted to extend the STI to predict the intelligibility of nonlinearly processed speech by proposing variations that use speech as a probe signal. This work considers four previously proposed speech-based STI methods and four novel methods, studied under conditions of additive noise, reverberation, and two nonlinear operations (envelope thresholding and spectral subtraction). Analyzing intermediate metrics in the STI calculation reveals why some methods fail for nonlinear operations. Results indicate that none of the previously proposed methods is adequate for all of the conditions considered, while four proposed methods produce qualitatively reasonable results and warrant further study. The discussion considers the relevance of this work to predicting the intelligibility of cochlear-implant processed speech.

Modified LMS algorithms for speech processing with an adaptive noise canceller

A desired signal corrupted by additive noise can often be recovered by an adaptive noise canceller using the least mean squares (LMS) algorithm. A major disadvantage of the LMS algorithm is its excess mean-squared error, or misadjustment, which increases linearly with the desired signal power, This leads to degrading performance when the desired signal exhibits large power fluctuations and is a serious problem in many speech processing applications. This work considers two modified LMS algorithms, the weighted sum and sum methods, designed to solve this problem by reducing the size of the steps in the weight update equation when the desired signal is strong. The weighted sum method is derived from an optimal method (also developed in this work), which is not generally applicable because it requires quantities unavailable in a practical system. The previously proposed, but ad hoc, sum method is analyzed and compared to the weighted sum method. Analysis of the two modified LMS algorithms indicates that either one provides substantial improvements in the presence of strong desired signals and similar performance in the presence of weak desired signals, relative to the unmodified LMS algorithm. Computer simulations with both uncorrelated Gaussian noise and speech signals confirm the results of the analysis and demonstrate the effectiveness of the modified algorithms. The modified LMS algorithms are particularly suited for signals (such as speech) that exhibit large fluctuations in short-time power levels.

Adaptive beamforming for noise reduction

The invention provides an adaptive noise cancelling apparatus which operates to overcome a problem encountered in conventional noise cancelling circuitry when the signal-to-noise ratio at the sensor array is high--to wit, that the target signal is degraded, leading to poorer intelligibility. An apparatus constructed in accord with the invention selectively inhibits the adaptive filter from changing its filter values in these instances and, thereby, prevents it from generating a noise-approximating signal that will degrade the target component of the output signal.

Evaluation of feedback-reduction algorithms for hearing aids.

Three adaptive feedback-reduction algorithms were implemented in a laboratory-based digital hearing aid system and evaluated with dynamic feedback paths and hearing-impaired subjects. The evaluation included measurements of maximum stable gain and subjective quality ratings. The continuously adapting CNN algorithm (Closed-loop processing with No probe Noise) provided the best performance: 8.5 dB of added stable gain (ASG) relative to a reference algorithm averaged over all subjects, ears, and vent conditions. Two intermittently adapting algorithms, ONO (Open-loop with Noise when Oscillation detected) and ONQ (Open-loop with Noise when Quiet detected), provided an average of 5 dB of ASG. Subjects with more severe hearing losses received greater benefits: 13 dB average ASG for the CNN algorithm and 7-8 dB average ASG for the ONO and ONQ algorithms. These values are conservative estimates of ASG because the fitting procedure produced a frequency-gain characteristic that already included precautions against feedback. Speech quality ratings showed no substantial algorithm effect on pleasantness or intelligibility, although subjects informally expressed strong objections to the probe noise used by the ONO and ONQ algorithms. This objection was not reflected in the speech quality ratings because of limitations of the experimental procedure. The results clearly indicate that the CNN algorithm is the most promising choice for adaptive feedback reduction in hearing aids.

Hands-on learning in biomedical signal processing

For 20 years, the Harvard-MIT Division of Health Sciences and Technology (HST) has offered a graduate-level course on Biomedical Signal and Image Processing (HST582J). This course takes a practical, hands-on approach to learning about signal processing and physiological signals through the application of digital signal processing methods to biomedical problems. It is by all accounts a successful course, with steadily increasing enrollment and high student satisfaction. In recent years, we have instituted a number of changes in the course, motivated by our awareness of new pedagogical strategies and assessment techniques. The purpose of this article is to use HST582J as a case study, demonstrating how incremental changes can improve an already successful course and ensure that hands-on exercises are not only fun and interesting but also achieve the desired learning objectives.

Microphone-Array Hearing Aids

Microphone-array hearing aids provide a promising solution to the problems encountered by hearing-impaired persons when listening to speech in the presence of background noise. This chapter first discusses implementation issues and performance metrics specific to the hearing-aid application. A review of previous work on microphone-array hearing aids includes systems with directional microphones, fixed beamformers, adaptive beamformers, physiologically-motivated processing, and binaural outputs. Recent simulation results of one promising adaptive beamforming system are presented. The performance of microphone-array hearing aids depends heavily on the acoustic environments in which they are used. Additional information about the level of reverberation, number of interferers, and relative levels of interferers encountered by hearing-aid users in everyday situations is required to quantify the benefit of microphone-array hearing aids and to select the optimal processing strategy.

Evaluation of array-processing algorithms for a headband hearing aid.

Several array-processing algorithms were implemented and evaluated with experienced hearing-aid users. The array consisted of four directional microphones mounted broadside on a headband worn on the top of the listeners head. The algorithms included two adaptive array-processing algorithms, one fixed array-processing algorithm, and a reference condition consisting of binaural directional microphones. The algorithms were evaluated under conditions with both one and three independent noise sources. Performance metrics included quantitative speech reception thresholds and qualitative subject preference ratings for ease-of-listening measured using a paired-comparison procedure. On average, the fixed algorithm improved speech reception thresholds by 2 dB, while the adaptive algorithms provided 7-9-dB improvement over the reference condition. Subjects judging ease-of-listening generally preferred all array-processing algorithms over the reference condition. The results suggest that these adaptive algorithms should be evaluated further in more realistic acoustic environments.

Sensitivity to design parameters in an adaptive-beamforming hearing aid

The optimal performance of a two-microphone adaptive-beamforming hearing aid was analyzed under sample conditions. Using a straight-ahead target and a single off-axis jammer, source-to-microphone impulse responses were recorded from head-mounted arrays in anechoic and reverberant environments. Optimal filters (equivalent to the fully converged beamformers adaptive filter) and associated system performance were computed for various filter structures and input target-to-jammer ratios. Generally, performance is sensitive to variations in filter length and filter primary-channel delay only under conditions of minimal reverberation and low target-to-jammer ratio. With realistic amounts of either reverberation or target, there is little sensitivity to filter length or primary channel delay.<<ETX>>