Patrick M. Peterson

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.

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.

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>>

Using linearly-constrained adaptive beamforming to reduce interference in hearing aids from competing talkers in reverberant rooms

Adaptive beamforming techniques can be applied to the problem of reducing interference in monaural hearing aids from sound sources that are spatially separated from a target source. We implemented a two-microphone Griffiths-Jim [1] beamformer in simulated reverberant environments (anechoic space, living room, and conference room) in which target sentences originated straight-ahead of the array and interfering speech babble originated 45° off-axis. We evaluated the system with intelligibility tests administered to normal-hearing subjects. Compared to listening through a single microphone, the two-microphone beamformer reduced the target-to-interference ratio required for 50% keyword intelligibility by 30, 14, and 0 dB in the anechoic, living-room, and conference-room conditions, respectively.

Reducing the effects of target misalignment in an adaptive beamformer for hearing aids

Previous work [Peterson et al., Acta Otolaryngol. (in press)] has shown that a multimicrophone adaptive beamforming system for hearing aids reduced noise effectively at input target‐to‐jammer ratios (TJRs) up to about 0 dB. However, that beamforming algorithm assumed that the target signal was identical at the microphones, and violation of this assumption caused degrading target cancellation when input TJR was greater than 0 dB. The present system uses a measure of intermicrophone correlation to estimate short‐time TJR and inhibits adaptation dynamically when TJR exceeds a selected threshold. Computer simulations of a two‐microphone system in anechoic and moderately reverberant environments show that this method reduces (speech) target cancellation when the identical target assumption is violated. The improved system reduces noise at both positive and negative input TJRs (±20 dB), with target misalignments up to 10 deg. Additionally, this processing scheme reduces the misadjustment noise caused by strong ...

An interactive environment for signal processing on a VAX computer

We have developed a software package, ISPUD, which has proven useful for signal processing. The package is based on the principles of data abstraction and represents signals as high-level data objects. It uses a simple interpreter to provide an easy-to-use, interactive, and extensible processing environment. In contrast to other abstraction-based packages, which require single-user LISP machines, ISPUD runs on a multi-user VAX with graphics terminals.

ISPUD, an interactive environment for signal processing on a VAX computer

To support our groups research in psychoacoustics and hearing aids, a useful software package, ISPUD, for computer‐based signal processing has been developed. Following the work of Kopec [IEEE Trans. Acoust. Speech, Signal Process. ASSP‐32, 842–851 (1984)], ISPUD uses abstract data objects to represent signals in a way that leads to a simple, interactive, and extensible environment for signal processing. In contrast to Kopecs package, which requires a LISP machine and supports one user, ISPUD runs on a VAX with the multi‐user VMS operating system, a C compiler, and Tektronix‐4010 compatible terminals. The user interface is a simple, LISP‐like interpreter for calling defined functions or combining functions into new definitions. The package includes efficient built‐in functions, written in C, for signal generation, arithmetic, composition, windowing, filtering, transforms, statistics, graphical display, and file I/O. In addition, user‐written C functions can be incorporated. Examples of ISPUD usage, incl...

Using multiple microphones and LMS adaptive beamforming to reduce interference in hearing aids from competing talkers in reverberant rooms

To reduce interference in monoaural hearing aids from spatially separated sound sources, the information from multiple microphones might be useful in enhancing the signal from a desired source for monaural presentation. The technique of Griffiths and Jim [IEEE Trans. Antennas Propagat. AP‐30, 27–34 (1982)] was used to filter and combine microphone signals ins way that preserves signals arriving from straight ahead of the microphone array while minimizing output power from off‐axis signals. The filters adapt to the interference environment using a variation of the LMS (Widrow) algorithm. Although it adapts slowly (1 s), LMS approaches optimum performance in stationary, anechoic environments, providing a bound on performance in nonstationary, reverberant environments. In three simulated rooms, anechoic, living room, and conference room, with interferring babble 45° off‐axis and on‐axis target at − 12 dB/S/B, a two‐microphone system with 100‐point filters reduced the interference by 20, 12, and 4 dB, respect...

Hearing‐aid microphone systems with increased directionality

Under some conditions of background noise and/or reverberation, increasing the directionality of hearing‐aid microphone systems can lead to gains in speech intelligibility. Conceptually, directional microphone systems are realized by combining the outputs of multiple, spatially distributed omnidirectional sensors. Two approaches to obtaining increased directivity D derive from the classical antenna theory: (1) With uniform or tapered sensor weightings, increasing the spatial extent L of the sensor array increases D. The maximum D is limited to 2Lf/c, where f is frequency and c is the speed of sound. (2) With arbitrarily small arrays, broadband “superdirectivity” can be achieved whereby the maximum D is independent of frequency and equals the square of the number of sensors in the array. A conventional hearing‐aid directional microphone exploits suboptimum superdirectivity; two closely spaced sensors attain D = 4.5 dB up to about 4 kHz. We shall review the basic principles and limitations of such microphone systems and report on our effort to develop a four‐sensor array (using two conventional directional microphones) that achieves D = 8.5 dB and a 3‐dB beamwidth of ± 37° for f < 4 kHz with an array length L = 4.1 cm that is practical for hearing‐aid applications. [Work supported by NIH.]

Evaluation of a two‐microphone speech‐enhancement system

Two variants of a processing scheme first proposed by Kaiser and David [J. Acoust. Soc. Am. 32, 918 (1960)] were evaluated for enhanced intelligibility. In both systems the running cross correlation between two microphone signals is used to derive a time‐varying gating signal that multiplies the sum of the microphone signals to form a monaural stimulus. With a target source straight ahead and an interference source off axis, this processing results in an output that is increasingly attenuated with decreasing target‐to‐interference ratio. In the first system this processing was applied to the wideband inputs and in the second it was applied independently in four octave bands. With the latter system three degrees of processing (determined by the function relating output attenuation to cross‐correlation) were investigated. Intelligibility tests were conducted using sentences as the straight‐ahead target and continuous discourse as the off‐axis (90° azimuth) interference. Contrary to expectations based on the...