Starkey Laboratories

An ear-wearable device includes at least two microphones arranged to provide separate directional information from ambient sound that has a speech component. A processor is configured to receive at least two audio signals from the respective at least two microphones. The at least two audio signals are mapped to an input layer of a deep neural network. The deep neural network is trained to estimate filter weights for an optimal directionality pattern which aims to extract the speech component from the ambient sound. The processor beamforms the at least two microphones to utilize the optimal directionality pattern.

An ear-wearable device includes a receiver that outputs an acoustic signal in response to a drive signal, a source device that provides a reference audio signal, and an inward facing microphone that converts in-ear acoustic sound to an error signal. The device includes an active noise controller that provides a noise control signal based on the error signal and a processing block that provides an amplified hearing signal based on the reference audio signal. The device includes an instability detector configured to: determine a transform function that represents an approximation of an inverse of the active noise controller; apply the transform function to the drive signal to determine an inverse filter signal; apply a filter to the error signal. The filter compensates for deficiencies in the transform function. The active noise controller is disabled when an anomalous similarity between filtered error signal and the inverse filter signal is detected.

Enhancing sound in an ear-wearable device involves receiving an ambient sound stream from a microphone of an ear-wearable device and inputting a representation of the ambient sound stream to a deep neural network that is trained to detect speech in a noisy environment. In response to detecting the speech via the deep neural network, the ear-wearable device operates in a directional mode in which an audio representation derived from the ambient sound stream is emphasized in a target direction. In response to detecting no speech via the deep neural network, the ear-wearable device operates in a spatial omnidirectional mode in which a fixed directional mode signal is combined with an omnidirectional signal to provide the audio representation. The fixed directional mode signal and the omnidirectional signal operate over different frequency ranges.

A system includes a hearing instrument and a vision device configured to be worn concurrently by a user. The hearing instrument receives an indication of a direction from the vision device. One or more processors of the hearing instrument receive first audio data representing sound detected by one or more microphones. The processors generate second audio data based on the first audio data, representing a version of the detected sound in which portions arriving from the indicated direction are enhanced. A receiver generates output sound based on the second audio data.

An effective howling prevention algorithm may provide improved sound quality of active noise cancellation (ANC) in audio amplification systems, such as in hearing assistance devices. A hearing device includes an inward-facing microphone and receiver, with a controller configured to monitor coherence between microphone and feedback signals to detect potential howling. When howling is detected based on an indicator of instability exceeding a threshold, the controller rapidly converges the gain to zero to prevent feedback howling. After a predetermined delay period without detected howling, the controller slowly converges the gain back to unity to restore active noise cancellation. This approach provides advantages over directly equating coherence to gain by more effectively preventing howling while maintaining optimal noise cancellation performance when stable.

This disclosure describes a hearing instrument comprising: a housing, a light source, a receiver, and one or more processors configured to: receive a device-finding signal from an external device, wherein the device-finding signal instructs the hearing instrument to enter a device-finding mode to assist a user in finding the hearing instrument; and in response to receiving the device-finding signal, cause the light source to emit a light signal; determine whether the hearing instrument is currently being worn; cause the receiver to output a first sound based on the hearing instrument currently being worn; and cause the receiver to output a second sound based on the hearing instrument not currently being worn.

Self-learning hearing device may enhance audio input based on an adaptive model. The adaptive, or changeable, model may be updatable in response to, among other things, user settings adjustments. For example, when a user makes a settings adjustment, the context surrounding, or corresponding to, the settings adjustment may be used to update the adaptive model.

A method comprises causing, by one or more processors of a hearing instrument, a receiver of the hearing instrument to output a sound; and causing, by the one or more processors, a light source of the hearing instrument to emit a first light signal indicating that the receiver output the sound.

An antenna for a hearing instrument comprises a first antenna segment; and a second antenna segment, wherein the first antenna segment is between a feed point of the antenna and the second antenna segment, the first antenna segment provides an in-line filter that is configured to attenuate one or more harmonic frequencies from an electrical current passing between the feed point and the second antenna segment via the first antenna segment.

Embodiments are directed to measuring, on a periodic basis by a hearing device, one or more transfer functions, each of the one or more transfer functions comprising a receiver-to-microphone acoustic transfer function of the hearing device. A plurality of health indicators of the hearing device are calculated over time using a measured characteristic of each of the measured transfer functions and a nominal characteristic of each of the measured transfer functions. Using the plurality of health indicators, degradation or failure of any of the receivers, any of the microphones, or a component in an acoustic path between one or more receivers and one or more microphones is detected. One or both of recommended actions and warnings are generated in response to detecting degradation or failure of any of the receivers, any of the microphones, or the component.