This invention concerns a unique passive headset to be worn by a user that wishes to block out background noise yet must hear speech. In industry there is a demand for noise canceling headsets requiring attenuation of noise in the speech band (300 to 3000 Hz) and above. Traditionally passive headsets work well at frequencies above 500 Hz but, unfortunately, attenuate speech. This invention provides a solution that provides protection offered by a passive headset but allows speech to pass via an external microphone through a speech filter and output to the headset loudspeaker.

The instant invention can operate with the algorithm described in U.S. Patent 5, 105, 377 which is hereby incorporated by reference herein.

Additionally, it can employ other algorithms such as that shown and described in U.S. Patent Application Ser. No. 07/421,759, to Ziegler et al and hereby incorporated by reference herein which is an "in- wire" selective cancellation system. U.S. Patents 4,025,721 and 4,185,168, to Graupe show systems that have attempted to combine high and low frequency range protection resulting in attenuation of the speech as well. The success of Graupe was only with fixed or "near stationary" noise but not with other noise of varying spectral characteristics and of brief duration. Accordingly it is an object of this invention to provide a passive selective headset that will attenuate unwanted background noise, thus allowing speech to pass. Another object of this invention is to provide a passive headset with a speech filter to allow speech in while blocking background noise.

A further object of this invention is to provide a passive headset which uses an "in wire" control algorithm to cancel only the noise thus allowing speech to pass with minimal degradation.

It is a further object of this invention to provide the option of manual selection of either speech filtering or "in-wire" control based upon the noise characteristics.

It is also an object of this invention to provide automatic selection of either speech filtering or "in-wire" control based upon the noise characteristics using a detection algorithm.

These and other objects will become apparent when reference is had to the accompanying drawing.

Fig. 1 is a diagrammatic view of the system comprising the instant invention. Fig. 2 is a more detailed description of the adaptive speech filtering implementation of Figure 1.

Fig. 3 is a more detailed description of the in-wire implementation of Figure 1. Fig. 4 is a diagrammatic view of the system shown in Figure 1 that incorporates a means for manually or automatically (via a detection algorithm) selecting the optional approach for passing speech.

The passive selective headset system is shown in Fig. 1 as 10. It consists of a typical passive headset 11, loudspeakers 12 that pass speech, reference microphones 14 that pick up exterior sound, i.e., speech, noise, warning signals, and a system controller 15 having either a speech filter 16 that separates speech from noise coming into the microphone or an in-wire controller 17 that attenuates only the noise coming into the microphone, thus allowing speech to pass.

Headset 11 has closed backs as at 16 for passive attenuation. Speech filter 16 or in-wire controller 17 is intended to separate speech from noise or attenuate only the noise, respectively, and output speech to the speakers 12 with a minimal effect on intelligibility. The specific approach to be used will be dependent on the type of noise that is to be controlled. In some cases, the filtering techniques of Graupe can be employed, in other cases, an "in-wire" control algorithm, e.g., digital virtual earth, can be used.

The headset speakers 12 need to be capable of reproducing the speech with little or no distortion and have a minimum of input-to-output delay so that speech is perceived to occur almost instantaneously.

The reference microphones 14 are typically small electret microphones that are attached to the outside of the headset at a distance from the ear canal. The reference microphone is used to allow outside speech and noise to enter the speech filter 16 or in- wire controller 17. The reference microphone 14 not only picks up damaging noise but also picks up speech and warning signals. The passive selective headset 11 is designed to provide protection against damaging noise above 500 Hz which typically also interferes with the speech band. It also attenuates noise in the audible range of the human ear effectively attenuating any warning siren. Using the reference microphones 14 as inputs to speech filter 16 or in-wire controller 17, or both, that is sent to speakers 12 is a means of allowing speech to pass through to the worker. If warning signals, common in heavy industrial settings, are at a known frequency, as most are, they can be filtered from the noise by filter 18 and passed to the worker as well. Fig. 1 shows this integration of passive hearing protection with speech and warning signal filtering so that local communication can be maintained for the comfort and safety of the worker. The block diagram shown in Figure 2 is a more detailed description of the speech filter implementation. The output, y^, of the filter is given by yk = kxk-l where x^.j is a delayed input signal comprised of speech and noise. Sk is a vector of filter coefficients that are set by estimates of the power of the noise, n^.

The input sample, x^ , is picked up by a reference sensor, passes through an amplifier, anti-aliasing filters, and then sampled at a high enough rate, to include the speech band, i.e. 300-3300 Hz before entering the adaptive speech filter process. The output speech sample, y^, is input to a D/A, reconstruction filters and an amplifier before driving the actuator with the speech signal.

The block diagram shown in Figure 3 is a more detailed description using an adaptive noise canceller such as a digital virtual earth control implementation. The output sample, y^ of the in-wire controller is given by

ⁿ k = k yk-l y _k = x _k + b _k where x _k is the input speech and noise signal. b _k is the cancellation output sample, i.e. the "anti-noise" sample. Λ _k is a vector of cancellation filter coefficients.

The input sample, x _k , is comprised of speech plus noise that is picked up by reference sensor, amplified, filtered, and sampled at a rate high enough to effectively control the noise to be canceled before input to the in-wire controller. The output sample, y _k , passes a D/A, filter and an amplifier before its "noise-less" signal drives the actuator.

In practice the digital virtual earth algorithm effectively separated siren noise from speech in an in-wire implementation for the passive selective headset. The adaptive speech filtering techniques of Graupe have been demonstrated to separate speech from hydrogen compressor noise or any stationary noise dramatically increasing the intelligibility of speech.

The implementation 40 shown in Figure 4 utilizes a manual switch or automatic detection algorithm 41 to select the most appropriate approach, i.e., speech filtering techniques or "in-wire" control techniques, or both, to most effectively pass speech. As stated, the speech filtering technique works extremely well on near-stationary noise and the "in-wire" control approaches perform effectively with the cancellation of varying tonal noise. Manual selection of the optimal approach can be accomplished via a switch. Automatic selection of an optimal approach can be accomplished by the use of a detection algorithm that measures the spectral characteristics, i.e., stationarity and amplitude, and selects a preferred approach, i.e., adaptive speech filtering or "in- wire" control, or both, based on the detected characteristics. The remainder of Figure 4 is identical to Figure 1.