Enhanced perception of target signals

Abstract

Presented herein are techniques for use of electrode-neural interface (ENI) factors to improve perception of target signals (e.g., speech) in noisy conditions. Some example techniques presented herein use signal-to-noise ratio (SNR) measures to improve speech-in- noise perception by incorporating ENI factors for monopolar stimulation strategies and / or focused stimulation strategies.

Description

Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 ENHANCED PERCEPTION OF TARGET SIGNALS BACKGROUND Field of the Invention

[0001] The present invention relates generally to electrically-stimulating devices and techniques for improving perception of target signals (e.g., speech) in noise by incorporating electrode neural interface factors for stimulation channel selection. Related Art

[0002] Medical devices have provided a wide range of therapeutic benefits to recipients over recent decades. Medical devices can include internal or implantable components / devices, external or wearable components / devices, or combinations thereof (e.g., a device having an external component communicating with an implantable component). Medical devices, such as traditional hearing aids, partially or fully-implantable hearing prostheses (e.g., bone conduction devices, mechanical stimulators, cochlear implants, etc.), pacemakers, defibrillators, functional electrical stimulation devices, and other medical devices, have been successful in performing lifesaving and / or lifestyle enhancement functions and / or recipient monitoring for a number of years.

[0003] The types of medical devices and the ranges of functions performed thereby have increased over the years. For example, many medical devices, sometimes referred to as “implantable medical devices,” now often include one or more instruments, apparatus, sensors, processors, controllers or other functional mechanical or electrical components that are permanently or temporarily implanted in a recipient. These functional devices are typically used to diagnose, prevent, monitor, treat, or manage a disease / injury or symptom thereof, or to investigate, replace or modify the anatomy or a physiological process. Many of these functional devices utilize power and / or data received from external devices that are part of, or operate in conjunction with, implantable components. SUMMARY

[0004] In one aspect, a method is provided. The first method comprises: determining channelized signals from an input signal; estimating channel signal-to-noise ratios (SNRs) for a plurality of the channelized signals; determining current spread functions for at least a subset of the plurality of channelized signals; estimating, based on the current spread functions and the channel SNRs, electrode-neural interface (ENI) signal-to-noise ratios (SNRs) for at leastAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 the subset of the plurality of channelized signals; and selecting, based on the ENI SNRs, a set of the channelized signals for use in delivering stimulation signals to a recipient.

[0005] In another aspect, processing unit for processing a spatial signal is provided. The processing unit for comprises: a filter bank configured to process the spatial signal to generate channel signals in each of a plurality of spaced frequency channels, wherein each channel is associated with at least one electrode of a plurality of electrodes configured to be implanted in a recipient; a channel signal-to-noise ratio (SNR) estimator configured to estimate channel SNRs associated with the channel signals; and an electrode-neural interface (ENI)-based channel selection module configured to: calculate, a vector of SNRs at an interface between one or more of the plurality of electrodes and neurons based on the channel SNRs, and select, based at least in part on the vector SNRs at the interface between the one or more of the plurality of electrodes and the neurons, a subset of the channel signals for use in stimulating the recipient based on the spatial signal.

[0006] In another aspect, a method is provided. The method comprises: obtaining a plurality of channel signals corresponding to an environmental signal; estimating local channel signal- to-noise ratios (SNRs) for each of the plurality of channel signals; calculating cumulative SNRs at an electrode-neural interface (ENI) for different subsets of channel signals among the plurality of channel signals based on the local channel SNRs; and selecting one of the different subsets of the plurality of channel signals for use in stimulation at least based on the cumulative SNRs at the ENI. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Embodiments of the present invention are described herein in conjunction with the accompanying drawings, in which:

[0008] FIG. 1A is a schematic diagram illustrating a cochlear implant system with which aspects of the techniques presented herein can be implemented;

[0009] FIG. 1B is a side view of a recipient wearing a sound processing unit of the cochlear implant system of FIG.1A;

[0010] FIG.1C is a schematic view of components of the cochlear implant system of FIG.1A;

[0011] FIG.1D is a block diagram of the cochlear implant system of FIG.1A;

[0012] FIG. 1E is a schematic diagram illustrating a computing device with which aspects of the techniques presented herein can be implemented;Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1

[0013] FIG. 2A is a schematic diagram illustrating a signal processing path of a cochlear implant system, including an ENI-based channel selection module, in accordance with embodiments presented herein;

[0014] FIG. 2B is a flowchart of an example method, in accordance with embodiments presented herein;

[0015] FIGs. 3A and 3B are stimulation diagrams illustrating a technique for improving effective SNR at the electrode-neural interface (ENI), in accordance with embodiments presented herein;

[0016] FIGs. 4A and 4B are stimulation diagrams illustrating a technique for improving channel independence based on SNR, in accordance with embodiments presented herein;

[0017] FIGs. 5A and 5B are stimulation diagrams illustrating a technique for improving channel independence and effective SNR at the ENI based on channel SNRs, in accordance with embodiments presented herein;

[0018] FIG.6A is a functional block diagram of a stimulation strategy adaptation module of a stimulation system, which includes a channel SNR estimator and an ENI-based channel selection module, in accordance with embodiments presented herein;

[0019] FIG. 6B is a flowchart of an example method, in accordance with embodiments presented herein;

[0020] FIG. 7A is a graph showing a standard spread of excitation, and a greater spread of excitation for a channel with a high SNR value, according to embodiments presented herein;

[0021] FIG.7B is an example of an SNR-based SPACE sound coding strategy (SNR-SPACE), according to embodiments presented herein;

[0022] FIG. 8A is a functional block diagram illustrating integration of a machine-learning stimulation module within a stimulation system, in accordance with certain embodiments presented herein; and

[0023] FIG. 8B is a functional block diagram of the machine-learning stimulation module of FIG.8A, in accordance with certain embodiments presented herein. DETAILED DESCRIPTION OverviewAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1

[0024] Cochlear implant recipients can have a difficult time understanding speech in noise. Noise reduction algorithms, while being somewhat effective, are usually focused on exploiting the characteristics of speech and noise, respectively. Existing noise reduction algorithms do not exploit any of the electrode-neural interface (ENI) factors that could provide additional gains to the recipient. To address these and other needs, presented herein are techniques for use of electrode-neural interface (ENI) factors to improve perception of speech in noisy conditions. Some example techniques presented herein use signal-to-noise ratio (SNR) measures to improve speech-in-noise perception by incorporating ENI factors for monopolar stimulation strategies and / or focused stimulation strategies. Some examples incorporate machine learning technologies for signal analysis, ENI-based SNR estimation, and stimulation control. Some example embodiments combine optimization of the front-end signal processing with optimization of the back-end electrical stimulation parameters, by optimizing channel / electrode selection based on including the full end-to-end system, instead of focusing individually on front-end signal processing, maxima selection, or the electrode-neural interface parts of the system. To that end, a method is provided to calculate not only per-channel SNR values, but also to calculate an effective / overall / cumulative (hereinafter, “cumulative”) SNR at the electrode-neural interface (also referred to herein as an “ENI SNR” value) based on the per-channel SNR values, in order to optimize channel / electrode selection for the delivery of stimulation signals to the recipient.

[0025] Using an amplitude-based electrode / channel selection technique, a certain set of channels / electrodes with the largest / highest amplitude will be selected for stimulation. However, using the SNR-based channel / electrode selection technique described herein forces a different set of channels / electrodes with better SNR to be selected for stimulation, depending on the contribution of signal / noise to the measured spread of excitation (SOE). Example embodiments can involve calculating a vector of ENI SNRs (at the neural elements), and then selecting an optimal set of channels / electrodes that will maximize the SNR at the electrode- neural interface, and thereby reduce, limit, or minimize the amount of interference / noise when delivering the stimulation signals to the recipient. That is, the techniques described herein optimize for the cumulative SNR observed at the neural elements for a given current spread (SOE), and then select the appropriate maxima (channels / electrodes). Depending on the signal and noise, the ENI SNR value can change even if the combined spectrum (of signal plus noise) is the same.Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1

[0026] There are a number of different types of devices in / with which embodiments of the present invention may be implemented. Merely for ease of description, the techniques presented herein are primarily described with reference to a specific device in the form of a cochlear implant system. However, it is to be appreciated that the techniques presented herein may also be partially or fully implemented by any of a number of different types of devices, including hearing devices, implantable medical devices, consumer electronic devices (e.g., mobile phones, tablets), wearable devices (e.g., smart watches), etc. As used herein, the term “hearing device” is to be broadly construed as any device that delivers sound signals to a user in any form, including in the form of acoustical stimulation, mechanical stimulation, electrical stimulation, etc. As such, a hearing device can be a device for use by a hearing-impaired person (e.g., hearing aids, middle ear auditory prostheses, bone conduction devices, direct acoustic stimulators, electro-acoustic hearing prostheses, auditory brainstem stimulators, bimodal hearing prostheses, bilateral hearing prostheses, dedicated tinnitus therapy devices, tinnitus therapy device systems, combinations or variations thereof, etc.) or a device for use by a person with normal hearing (e.g., consumer devices that provide audio streaming, consumer headphones, earphones and other listening devices). In other examples, the techniques presented herein can be implemented by, or used in conjunction with, various implantable medical devices, such as vestibular devices (e.g., vestibular implants), visual devices (i.e., bionic eyes), sensors, pacemakers, drug delivery systems, defibrillators, functional electrical stimulation devices, catheters, seizure devices (e.g., devices for monitoring and / or treating epileptic events), sleep apnea devices, electroporation devices, etc.

[0027] In addition, for ease of description, the techniques presented herein are primarily described with reference to input signals in the form of sound signals. However, it is to be appreciated that the techniques presented herein can be applied to processing of other types of sound signals, such as spatial signals or environmental sound signals. Examples of spatial signals can include, for example, sound signals, light or image signals, motion / balance signals, etc. Example System, Devices, and Components

[0028] FIGs.1A-1D illustrates an example cochlear implant system 102 with which aspects of the techniques presented herein can be implemented. The cochlear implant system 102 comprises an external component 104 that is configured to be directly or indirectly attached to the body of the user, and an internal / implantable component 112 that is configured to be implanted in or worn on the head of the user. In the examples of FIGs.1A-1D, the implantableAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 component 112 is sometimes referred to as a “cochlear implant.” FIG. 1A illustrates the cochlear implant 112 implanted in the head 154 of a user, while FIG.1B is a schematic drawing of the external component 104 worn on the head 154 of the user. FIG.1C is another schematic view of the cochlear implant system 102, while FIG. 1D illustrates further details of the cochlear implant system 102. For ease of description, FIGs.1A-1D will generally be described together.

[0029] In the examples of FIGs. 1A-1D, the external component 104 comprises a sound processing unit 106, an external coil 108, and generally, a magnet fixed relative to the external coil 108. The cochlear implant 112 includes an implantable coil 114, an implant body 134, and an elongate stimulating assembly 116 configured to be implanted in the user’s cochlea. In one example, the sound processing unit 106 is an off-the-ear (OTE) sound processing unit, sometimes referred to herein as an OTE component, that is configured to send data and power to the implantable component 112. In general, an OTE sound processing unit is a component having a generally cylindrically shaped housing 111 and which is configured to be magnetically coupled to the user’s head 154 (e.g., includes an integrated external magnet 150 configured to be magnetically coupled to an internal / implantable magnet 152 in the implantable component 112). The OTE sound processing unit 106 also includes an integrated external (headpiece) coil 108 (the external coil 108) that is configured to be inductively coupled to the implantable coil 114.

[0030] It is to be appreciated that the OTE sound processing unit 106 is merely illustrative of the external devices that could operate with implantable component 112. For example, in alternative examples, the external component 104 may comprise a behind-the-ear (BTE) sound processing unit configured to be attached to, and worn adjacent to, the recipient’s ear. In general, a BTE sound processing unit comprises a housing that is shaped to be worn on the outer ear of the user and is connected to the separate external coil assembly via a cable, where the external coil assembly is configured to be magnetically and inductively coupled to the implantable coil 114. It is also to be appreciated that alternative external components could be located in the user’s ear canal, worn on the body, etc.

[0031] Although the cochlear implant system 102 includes the sound processing unit 106 and the cochlear implant 112, as described below, the cochlear implant 112 can operate independently from the sound processing unit 106, for at least a period, to stimulate the user. For example, the cochlear implant 112 can operate in a first general mode, sometimes referred to as an “external hearing mode,” in which the sound processing unit 106 captures soundAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 signals which are then used as the basis for delivering stimulation signals to the user. The cochlear implant 112 can also operate in a second general mode, sometimes referred as an “invisible hearing” mode, in which the sound processing unit 106 is unable to provide sound signals to the cochlear implant 112 (e.g., the sound processing unit 106 is not present, the sound processing unit 106 is powered-off, the sound processing unit 106 is malfunctioning, etc.). As such, in the invisible hearing mode, the cochlear implant 112 captures sound signals itself via implantable sound sensors and then uses those sound signals as the basis for delivering stimulation signals to the user. Further details regarding operation of the cochlear implant 112 in the external hearing mode are provided below, followed by details regarding operation of the cochlear implant 112 in the invisible hearing mode. It is to be appreciated that reference to the external hearing mode and the invisible hearing mode is merely illustrative and that the cochlear implant 112 could also operate in alternative modes.

[0032] In FIGs.1A and 1C, the cochlear implant system 102 is shown with an external device 110, configured to implement aspects of the techniques presented. The external device 110, which is shown in greater detail in FIG.1E, is a computing device, such as a personal computer (e.g., laptop, desktop, tablet), a mobile phone (e.g., smartphone), remote control unit, etc. The external device 110 and the cochlear implant system 102 (e.g., sound processing unit 106 or the cochlear implant 112) wirelessly communicate via a bi-directional communication link 126. The bi-directional communication link 126 may comprise, for example, a short-range communication, such as Bluetooth link, Bluetooth Low Energy (BLE) link, a proprietary link, etc.

[0033] Returning to the example of FIGs.1A-1D, the sound processing unit 106 of the external component 104 also comprises one or more input devices configured to capture and / or receive input signals (e.g., sound or data signals) at the sound processing unit 106. The one or more input devices include, for example, one or more sound input devices 118 (e.g., one or more external microphones, audio input ports, telecoils, etc.), one or more auxiliary input devices 128 (e.g., audio ports, such as a Direct Audio Input (DAI), data ports, such as a Universal Serial Bus (USB) port, cable port, etc.), and a short-range wireless transmitter / receiver (wireless transceiver) 120 (e.g., for communication with the external device 110), each located in, on or near the sound processing unit 106. However, it is to be appreciated that one or more input devices may include additional types of input devices and / or less input devices (e.g., the short- range wireless transceiver 120 and / or one or more auxiliary input devices 128 could be omitted).Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1

[0034] The sound processing unit 106 also comprises the external coil 108, a charging coil 130, a closely-coupled radio frequency transmitter / receiver (RF transceiver) 122, at least one rechargeable battery 132, and an external sound processing module 124. The external sound processing module 124 can comprise various modules (logic) stored in a memory and executed by a processor. More specifically, the external sound processing module 124 can be configured to perform a number of operations which are represented in FIG.1D by sound processing logic 131 and ENI-based channel selection logic 162. Each of sound processing logic 131 and ENI- based channel selection logic 162 can be formed by one or more processors (e.g., one or more Digital Signal Processors (DSPs), one or more uC cores, etc.), firmware, software, etc. arranged to perform operations described herein. That is, sound processing logic 131 and ENI-based channel selection logic 162 can each be implemented as firmware elements, partially or fully implemented with digital logic gates in one or more application-specific integrated circuits (ASICs), partially or fully in software, etc. Although FIG.1D illustrates the ENI-based channel selection logic 162 as being implemented / performed at the external sound processing module 124, it is to be appreciated that this element (e.g., functional operations) could also or alternatively be implemented / performed as part of the implantable sound processing module 158, as part of the external device 110, etc. This flexibility regarding the implementation of the ENI-based channel selection logic 162 is indicated by dashed lines in FIG.1D. As further described below, certain aspects of the sound processing logic 131 and / or the ENI-based channel selection logic 162 can be machine-learned logic that is obtained by training one or more models to implement the techniques described herein using a large number of training samples involving various sound environments, speech conditions, noise conditions, speech- in-noise conditions, current functions (spread of excitation (SOE) patterns), signal-to-noise ratios (SNRs), electrode-neural interface (ENI) factors, physiological factors, user preferences, and the like.

[0035] Returning to the example of FIGs. 1A-1D, the implantable component 112 comprises an implant body (main module) 134, a lead region 136, and the intra-cochlear stimulating assembly 116, all configured to be implanted under the skin (tissue) 115 of the user. The implant body 134 generally comprises a hermetically-sealed housing 138 that includes, in certain examples, at least one power source 125 (e.g., one or more batteries, one or more capacitors, etc.) 125, in which RF interface circuitry 140 and a stimulator unit 142 are disposed. The implant body 134 also includes the internal / implantable coil 114 that is generally externalAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 to the housing 138, but which is connected to the RF interface circuitry 140 via a hermetic feedthrough (not shown in FIG.1D).

[0036] As noted, stimulating assembly 116 is configured to be at least partially implanted in the user’s cochlea. Stimulating assembly 116 includes a plurality of longitudinally spaced intra-cochlear electrical stimulating contacts (electrodes) 144 that collectively form a contact array (electrode array) 146 for delivery of electrical stimulation (current) to the recipient’s cochlea. Stimulating assembly 116 extends through an opening in the recipient’s cochlea (e.g., cochleostomy, the round window, etc.) and has a proximal end connected to stimulator unit 142 via lead region 136 and a hermetic feedthrough (not shown in FIG.1D). Lead region 136 includes a plurality of conductors (wires) that electrically couple the electrodes 144 to the stimulator unit 142. The implantable component 112 also includes an electrode outside of the cochlea, sometimes referred to as the extra-cochlear electrode (ECE) 139.

[0037] As noted, the cochlear implant system 102 includes the external coil 108 and the implantable coil 114. The external magnet 150 is fixed relative to the external coil 108 and the internal / implantable magnet 152 is fixed relative to the implantable coil 114. The external magnet 150 and the internal / implantable magnet 152 fixed relative to the external coil 108 and the internal / implantable coil 114, respectively, facilitate the operational alignment of the external coil 108 with the implantable coil 114. This operational alignment of the coils enables the external component 104 to transmit data and power to the implantable component 112 via a closely-coupled wireless link 148 formed between the external coil 108 with the implantable coil 114. In certain examples, the closely-coupled wireless link 148 is a radio frequency (RF) link. However, various other types of energy transfer, such as infrared (IR), electromagnetic, capacitive and inductive transfer, may be used to transfer the power and / or data from an external component to an implantable component and, as such, FIG. 1D illustrates only one example arrangement.

[0038] As noted above, sound processing unit 106 includes the external sound processing module 124. The external sound processing module 124 is configured to process the received input audio signals (received at one or more of the input devices, such as sound input devices 118 and / or auxiliary input devices 128), and convert the received input audio signals into output control signals for use in stimulating a first ear of a recipient or user (i.e., the external sound processing module 124 is configured to perform sound processing on input signals received at the sound processing unit 106). Stated differently, the one or more processors (e.g., processing element(s) implementing firmware, software, etc.) in the external sound processing moduleAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 124 are configured to execute sound processing logic in memory to convert the received input audio signals into output control signals (stimulation signals) that represent electrical stimulation for delivery to the recipient.

[0039] As noted, FIG. 1D illustrates an embodiment in which the external sound processing module 124 in the sound processing unit 106 generates the output control signals. In an alternative embodiment, the sound processing unit 106 can send less processed information (e.g., audio data) to the implantable component 112 and the sound processing operations (e.g., conversion of input sounds to output control signals 156) can be performed by a processor within the implantable component 112.

[0040] In FIG. 1D, according to an example embodiment, output control signals (stimulation signals) are provided to the RF transceiver 122, which transcutaneously transfers the output control signals (e.g., in an encoded manner) to the implantable component 112 via external coil 108 and implantable coil 114. That is, the output control signals (stimulation signals) are received at the RF interface circuitry 140 via implantable coil 114 and provided to the stimulator unit 142. The stimulator unit 142 is configured to utilize the output control signals to generate electrical stimulation signals (e.g., current signals) for delivery to the user’s cochlea via one or more of the stimulating contacts (electrodes) 144. In this way, cochlear implant system 102 electrically stimulates the user’s auditory nerve cells, bypassing absent or defective hair cells that normally transduce acoustic vibrations into neural activity, in a manner that causes the recipient to perceive one or more components of the input audio signals (the received sound signals).

[0041] As detailed above, in the external hearing mode the cochlear implant 112 receives processed sound signals from the sound processing unit 106. However, in the invisible hearing mode, the cochlear implant 112 is configured to capture and process sound signals for use in electrically stimulating the user’s auditory nerve cells. In particular, as shown in FIG.1D, an example embodiment of the cochlear implant 112 can include a plurality of implantable sound sensors 165(1), 165(2) that collectively form a sensor array 160, and an implantable sound processing module 158. Similar to the external sound processing module 124, the implantable sound processing module 158 may comprise, for example, one or more processors and a memory device (memory) that includes sound processing logic. The memory device may comprise any one or more of: Non-Volatile Memory (NVM), Ferroelectric Random Access Memory (FRAM), read only memory (ROM), random access memory (RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical,Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 or other physical / tangible memory storage devices. The one or more processors are, for example, microprocessors or microcontrollers that execute instructions for the sound processing logic stored in memory device.

[0042] In the invisible hearing mode, the implantable sound sensors 165(1), 165(2) of the sensor array 160 are configured to detect / capture input sound signals 166 (e.g., acoustic sound signals, vibrations, etc.), which are provided to the implantable sound processing module 158. The implantable sound processing module 158 is configured to convert received input sound signals 166 (received at one or more of the implantable sound sensors 165(1), 165(2)) into output control signals 156 for use in stimulating the first ear of a recipient or user (i.e., the implantable sound processing module 158 is configured to perform sound processing operations). Stated differently, the one or more processors (e.g., processing element(s) implementing firmware, software, etc.) in implantable sound processing module 158 are configured to execute sound processing logic in memory to convert the received input sound signals 166 into output control signals 156 that are provided to the stimulator unit 142. The stimulator unit 142 is configured to utilize the output control signals 156 to generate electrical stimulation signals (e.g., current signals) for delivery to the user’s cochlea, thereby bypassing the absent or defective hair cells that normally transduce acoustic vibrations into neural activity.

[0043] It is to be appreciated that the above description of the so-called external hearing mode and the so-called invisible hearing mode are merely illustrative and that the cochlear implant system 102 could operate differently in different embodiments. For example, in one alternative implementation of the external hearing mode, the cochlear implant 112 could use signals captured by the sound input devices 118 and the implantable sound sensors 165(1), 165(2) of sensor array 160 in generating stimulation signals for delivery to the user.

[0044] In the examples of FIGs. 1A-1D, certain aspects of the techniques presented herein, such as the sound processing logic 131 and / or the ENI-based channel selection logic 162, can be performed by one or more components of the cochlear implant system 102, such as the external sound processing module 124, the implantable sound processing module 158, an / or the external device 110, etc. This is generally shown by dashed boxes 162 in FIG.1D. That is, dashed boxes 162 generally represent potential locations for some or all of the ENI-based channel selection logic 162 that, when executed, is configured to perform aspects of the techniques presented herein. As noted above, the external sound processing module 124, the implantable sound processing module 158, and / or the external device 110 may comprise, forAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 example, one or more processors and a memory device (memory) that includes all or part of the sound processing logic and / or the ENI-based channel selection logic 162. The memory device may comprise any one or more of: non-volatile memory (NVM), random access memory (RAM), FRAM, read only memory (ROM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. Thus, in general, the memory may comprise one or more tangible (non-transitory) computer readable storage media (e.g., a memory device) encoded with software comprising computer executable instructions. The one or more processors are, for example, microprocessors or microcontrollers that execute the instructions for the sound processing logic 131 and / or the ENI-based channel selection logic 162 stored in the memory device. When the software is executed (by the one or more processors), it is operable to perform the operations described herein with reference to sound processing, channel SNR estimation, ENI-based SNR estimation, channel selection, stimulation strategy, sound coding, and the training and use of one or more machine learning models for estimating the ENI-based SNR values and setting / adapting the stimulation strategy (e.g., to dynamically perform channel selection according to electrode-neural interface (ENI) factors, such as the cumulative SNR at the ENI, as described herein). In some example embodiments, certain aspects of the sound processing logic 131 and / or the ENI-based channel selection logic 162 may be distributed between the external sound processing module 124, the implantable sound processing module 158, and the external device 110. Although software implementations for the external sound processing module 124 are described, one or more operations associated with the external sound processing module 124 and any of its logic components or modules can be partially or fully implemented with digital logic gates in one or more application- specific integrated circuits (ASICs).

[0045] A machine-learning stimulation device comprising ENI-based channel selection logic 162 is described as a functional block (e.g., one or more processors operating based on code, algorithm(s), etc.) that is trained, through a machine-learning process, to detect events, such as events relating to an external sound environment of a user (e.g., various speech-in-noise conditions, channel SNRs, ENI SNRs, etc.). A machine-learning stimulation device presented herein is further trained, via the same or different machine-learning process, to set / determine a stimulation strategy to be delivered to the user in response to the detected event, that accounts for attributes of the events. That is, the techniques presented herein use one or more machine- learning models to automatically estimate ENI SNR values (e.g., based on channel SNR valuesAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 and current spread functions) and select an optimal stimulation strategy to adapt to an automatically detected event (e.g., identify an optimal configuration of channels (i.e., a set of active stimulating electrodes) that would improve perception of speech in noise by the recipient according to the electrode-neural interface factors).

[0046] FIG. 1E is a block diagram illustrating one example arrangement for an external computing device 110 configured to perform one or more operations in accordance with certain embodiments presented herein. As shown in FIG. 1E, in its most basic configuration, the external computing device 110 includes at least one processing unit 183 and a memory 184. The processing unit 183 includes one or more hardware or software processors (e.g., Central Processing Units) that can obtain and execute instructions. The processing unit 183 can communicate with and control the performance of other components of the external computing device 110. The memory 184 is one or more software or hardware-based computer-readable storage media operable to store information accessible by the processing unit 183. The memory 184 can store, among other things, instructions executable by the processing unit 183 to implement applications or cause performance of operations described herein, as well as other data. The memory 184 can be volatile memory (e.g., RAM), non-volatile memory (e.g., ROM), or combinations thereof. The memory 184 can include transitory memory or non-transitory memory. The memory 184 can also include one or more removable or non-removable storage devices. In examples, the memory 184 can include random access memory (RAM), read only memory (ROM), EEPROM (Electronically-Erasable Programmable Read-Only Memory), flash memory, optical disc storage, magnetic storage, solid state storage, or any other memory media usable to store information for later access. By way of example, and not limitation, the memory 184 can include wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared and other wireless media or combinations thereof. In certain embodiments, the memory 184 comprises sound processing logic 131 and ENI-based channel selection logic 162 that, when executed, enables the processing unit 183 to perform aspects of the techniques presented. As stated, the sound processing logic 131 and / or the ENI- based channel selection logic 162 can be machine-learned by training one or more models using a large corpus of sample data to implement the techniques described herein.

[0047] In the illustrated example of FIG. 1E, the external computing device 110 further includes a network adapter 186, one or more input devices 187, and one or more output devices 188. The external computing device 110 can include other components, such as a system bus, component interfaces, a graphics system, a power source (e.g., a battery), among otherAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 components. The network adapter 186 is a component of the external computing device 110 that provides network access (e.g., access to at least one network 189). The network adapter 186 can provide wired or wireless network access and can support one or more of a variety of communication technologies and protocols, such as ETHERNET, cellular, BLUETOOTH, near-field communication, and RF (Radiofrequency), among others. The network adapter 186 can include one or more antennas and associated components configured for wireless communication according to one or more wireless communication technologies and protocols. The one or more input devices 187 are devices over which the external computing device 110 receives input from a user. The one or more input devices 187 can include physically- actuatable user-interface elements (e.g., buttons, switches, or dials), a keypad, keyboard, mouse, touchscreen, and voice input devices, among other input devices that can accept user input. The one or more output devices 188 are devices by which the computing device 110 is able to provide output to a user. The output devices 188 can include, a display 190 (e.g., a liquid crystal display (LCD)) and one or more speakers 191, among other output devices for presentation of visual or audible information to the recipient, a clinician, an audiologist, or other user.

[0048] It is to be appreciated that the arrangement for the external computing device 110 shown in FIG. 1E is merely illustrative and that aspects of the techniques presented herein can be implemented at a number of different types of systems / devices including any combination of hardware, software, and / or firmware configured to perform the functions described herein. For example, the external computing device 110 can be a personal computer (e.g., a desktop or laptop computer), a hand-held device (e.g., a tablet computer), a mobile device (e.g., a smartphone), a surgical system, and / or any other electronic device having the capabilities to perform the associated operations described elsewhere herein.

[0049] As described further below, electrical stimulation has a number of characteristics / attributes that control the stimulation strategy, including but not limited to the stimulus resolution, channel selection, etc. Control of the electrical stimulation width along the frequency axis (i.e., along the basilar membrane) of an area of activated nerve cells in response to delivered stimulation, sometimes referred to herein as the “spatial resolution” of the electrical stimulation. The spatial resolution of electrical stimulation may be controlled, for example, through the use of different electrode configurations for a given stimulation channel to activate nerve cell regions of different widths. Monopolar stimulation, for instance, is an electrode configuration where for a given stimulation channel the current is “sourced” viaAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 one of the intra-cochlea electrodes 144, but the current is “sunk” by an electrode outside of the cochlea, sometimes referred to as the extra-cochlear electrode (ECE) 139 (refer to FIG. 1D). Monopolar stimulation typically exhibits a large degree of current spread (i.e., wide stimulation pattern) and, accordingly, has a low spatial resolution. Other types of electrode configurations, such as bipolar, tripolar, focused multi-polar (FMP), a.k.a. “phased-array” stimulation, etc. typically reduce the size of an excited neural population by “sourcing” the current via one or more of the intra-cochlear electrodes 144, while also “sinking” the current via one or more other proximate intra-cochlear electrodes. Bipolar, tripolar, focused multi-polar and other types of electrode configurations that both source and sink current via intra-cochlear electrodes are generally and collectively referred to herein as “focused” stimulation. Focused stimulation typically exhibits a smaller degree of current spread (i.e., narrow stimulation pattern) when compared to monopolar stimulation and, accordingly, has a higher spatial resolution than monopolar stimulation. Likewise, other types of electrode configurations, such as double electrode mode, virtual channels, wide channels, defocused multi-polar, etc. typically increase the size of an excited neural population by “sourcing” the current via multiple neighboring intra-cochlear electrodes.

[0050] The cochlea is tonotopically mapped, that is, partitioned into regions each responsive to sound signals in a particular frequency range. In general, the basal region of the cochlea is responsive to higher frequency sounds, while the more apical regions of the cochlea are responsive to lower frequencies. The tonotopic nature of the cochlea is leveraged in cochlear implants such that specific acoustic frequencies are allocated to the electrodes 144 of the stimulating assembly that are positioned close to the corresponding tonotopic region of the cochlea (i.e., the region of the cochlea that would naturally be stimulated in acoustic hearing by the acoustic frequency). That is, in a cochlear implant, specific frequency bands are each mapped to a set of one or more electrodes that are used to stimulate a selected (target) population of cochlea nerve cells. The frequency bands and associated electrodes form a stimulation “channel” that delivers stimulation signals to the recipient.

[0051] Due to the tonotopic mapping of the cochlea, different portions of the sound signals are delivered to different target locations / places of the cochlea via different stimulation channels. As used herein, a stimulation channel is a combination / set of the electrodes that are used simultaneously / collectively to deliver current signals to the cochlea so as to elicit stimulation at a specific target location / place of the cochlea. For ease of illustration, stimulation channels are generally defined herein with reference to a corresponding “central electrode” and one orAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 more “secondary electrodes.” The central electrode of a stimulation channel is the electrode closest to the peak in the voltage (e.g., the electrode closest to the specific target location / place of the cochlea). The secondary electrodes of a stimulation channel are one or more electrodes other than the central electrode that are used to control the resulting voltage fields. Within a given stimulation channel, the current signals at different secondary electrodes may have different polarities (i.e., the current can be sourced at certain electrodes and sunk at other electrodes) and different relative magnitudes. The magnitudes of the current signals are generated in accordance with predetermined relative asymmetric current weights. Within a given stimulation channel, the current signals at the different electrodes have relative current weights so that the current signals evoke perception of a sound frequency portion at the target location of the cochlea. The process for determining the magnitudes of current signals for stimulation, given a predetermined set of current weights for the channel, is sometimes referred to herein as channel encoding. The channel encoding process may be implemented by the sound processing unit 106 (e.g., external sound processing module 124), or by the cochlear implant 112 (e.g., implantable sound processing module 158), or by a combination of these components of the cochlear implant system 102. For example, the sound processing unit 106 can be used to implement various sound encoding schemes, including but not limited to an Advanced Combination Encoder (ACE) sound encoding scheme, a Spread Pre-compensation Advanced Combination Encoder (SPACE) sound encoding scheme, and / or a signal-to-noise ratio based SPACE (SNR-SPACE) sound encoding scheme according to certain example embodiments described herein.

[0052] In addition, not all stimulation channels are activated (used to deliver stimulation signals) at a given time instances. The process for selecting which of the stimulation channels to use at a given time instance (e.g., to evoke perception of sounds received / captured within a time period) is referred to as “channel selection.” The channel selection process results in the activation of a subset of channels, sometimes referred to as N of M, at a given time instance.

[0053] FIG.2A is a schematic diagram illustrating a sound processing path 250 of a cochlear implant, such as cochlear implant system 102, in accordance with embodiments presented herein. As noted, the cochlear implant 102 comprises one or more sound input elements 118. In the example of FIG. 2A, the sound input elements 118 comprise two microphones 218. If not already in an electrical form, sound input elements 118 (microphones 218) convert received / input sound signals into electrical signals, referred to herein as electrical input signalsAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 251, that represent the received sound signals. As shown in FIG.2A, the electrical input signals 251 are provided to a pre-processing module 252.

[0054] The pre-processing module 252 is configured to combine the electrical input signals 251 received from the microphones 218, and prepare those signals for subsequent processing, as needed. The pre-processing module 252 then generates a pre-processed broadband signal 253 that, as described further below, is the basis of further processing operations. The pre- processed broadband signal 253 represents the collective sound signals received at the microphones 218 at a given point in time. The pre-processing module 252 can include or utilize a pre-emphasis filter, automatic gain control (AGC), manual sensitivity control (MSC), and / or noise reduction techniques, for example, during the front-end processing stage.

[0055] The cochlear implant 102 is generally configured to execute sound processing and coding to convert the pre-processed broadband signal 253 into output signals that represent electrical stimulation for delivery to the recipient. As such, the sound processing path 250 comprises a filter-bank module 254, a post-processing module 256, a channel SNR estimator module 258, and a stimulation strategy adaptation module 260. The stimulation strategy adaptation module 260 comprises an ENI-based channel selection module 262 and a mapping and encoding module 264.

[0056] In operation, the pre-processed broadband signal 253 generated by the pre-processing module 252 is provided to the filter-bank module 254. The filter-bank module 254 generates a suitable set of bandwidth limited channels, or frequency bins, that each includes a spectral component of the received sound signals. That is, the filter-bank module 254 comprises a plurality of band-pass filters that separate the pre-processed broadband signal 253 into multiple components / channels, each one carrying / representing a single frequency sub-band of the original signal (i.e., frequency components of the received sounds signal). In some examples, the filter-bank module 254 can comprise a set of band-pass filters with contiguous frequency boundaries, or can comprise a set of broader band-pass filters with overlapping frequency boundaries. The filter-bank module 254 can output ‘m’ band-pass filtered signal amplitudes, where ‘m’ is the total number of available stimulation channels for the cochlear implant. For an implant system providing 22 channels of stimulation, the filter-bank module 254 can output 22 separate band-pass filtered signals (22 channel signals), one for each stimulation channel. In some examples, the band-pass filtered signal amplitudes from the filter-bank module 254 can be used in selecting the stimulation channels (sets of electrodes) for application of stimulation and the amplitudes (current levels) for the applied stimulation.Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1

[0057] The channels created by the filter-bank module 254 are sometimes referred to herein as sound processing channels, and the sound signal components within each of the sound processing channels are sometimes referred to herein as “band-pass filtered signals” or “channelized signals.” The band-pass filtered or channelized signals created by the filter-bank module 254 are processed (e.g., modified / adjusted) as they pass through the sound processing path 250. As such, the band-pass filtered or channelized signals are referred to differently at different stages of the sound processing path 250. However, it will be appreciated that reference herein to a band-pass filtered signal or a channelized signal may refer to the spectral component of the received sound signals at any point within the sound processing path 250 (e.g., pre-processed, processed, selected, etc.).

[0058] At the output of the filter-bank module 254, the channelized signals are initially referred to herein as filtered channelized signals 255. The number ‘m’ of channels and filtered channelized signals 255 generated by the filter-bank module 254 may depend on a number of different factors including, but not limited to, implant design, number of active electrodes, coding strategy, and / or recipient preference(s). In certain arrangements, twenty-two (22) channelized signals are created and the sound processing path 250 is said to include 22 channels.

[0059] The filtered channelized signals 255 are provided to the post-processing module 256. The post-processing module 256 is configured to perform a number of sound processing operations on the filtered channelized signals 255. These sound processing operations can include, for example, channelized gain adjustments for hearing loss compensation (e.g., gain adjustments to one or more discrete frequency ranges of the sound signals), noise reduction operations, speech enhancement operations, envelope detection, etc., in one or more of the channels. After performing the sound processing operations, the post-processing module 256 outputs a plurality of post-processed channelized signals 257.

[0060] Also shown in FIG. 2A is a channel SNR estimator module 258 that is configured to evaluate / analyze the input sound signals and determine per-channel signal-to-noise ratios (channel SNRs) for each of the channelized signals. Optionally, the channel SNR estimator 258 could also determine the sound class of the sound signals by using the received sound signals to “classify” the ambient sound environment and / or the sound signals into one or more sound categories (i.e., to determine the input signal type). The sound classes / categories may include, but are not limited to, “Speech,” “Speech+Noise,” “Noise,” “Music,” and “Quiet.” In one non-limiting example, the channel SNR estimator 258 could use the determined soundAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 class in making a determination to estimate the channel SNRs of the sound signals. In one example, the operations of the channel SNR estimator 258 are performed using post-processed channelized signals 257 generated by the post-processing module 256 as input (represented by solid arrow 257 in FIG.2A). In another example, the operations of the channel SNR estimator 258 can be performed using the filtered channelized signals 255 that are output by the filter- bank module 254 as the input (represented by solid arrow 255 in FIG.2A).

[0061] In yet another example, the operations of the channel SNR estimator 258 could be performed using the pre-processed broadband signal 253 generated by the pre-processing module 252 as input (represented by dashed arrow 253 in FIG. 2A). In this further example, the channel SNR estimator 258 can also additionally calculate the global SNR of the broadband signal based on the pre-processed broadband signal 253 in addition to the narrow band channel SNRs (i.e., based on either the filtered channelized signals 255 or the post-processed channelized signals 257). The global SNR can also help with decisions about “Quiet,” “Speech,” or “Speech+Noise” more directly, for example.

[0062] The channel SNR estimator 258 generates channel SNR values 259 that are provided to the stimulation strategy adaptation module 260. Optionally, the channel SNR estimator 258 could also output sound classification data representing the sound class / category of the sound signals, along with the channel SNR values. Based on the post-processed channelized signals 257 (or the filtered channelized signals 255) and the channel SNR values 259, the stimulation strategy adaptation module 260 is configured to determine a stimulation strategy that should be used in delivering electrical stimulation signals to represent (evoke perception of) the sound signals. The stimulation strategy that should be used in delivering electrical stimulation signals is sometimes referred to herein as the “target” stimulation strategy.

[0063] As noted, the sound processing path 250 also includes the stimulation strategy adaptation module 260. In the specific arrangement of FIG. 2A, the stimulation strategy adaptation module 260 comprises an ENI-based channel selection module 262 and a mapping and encoding module 264. Although the channel SNR estimator 258 is shown as a separate module in FIG.2A, it should be appreciated that example embodiments are not limited thereto, and that the channel SNR estimator 258 could be integrated as part of the stimulation strategy adaptation module 260 or the ENI-based channel selection module 262 in some other example embodiments. The ENI-based channel selection module 262 is configured to perform a channel selection process to select, according to one or more selection rules, which subset of ‘n’ channels among the ‘m’ total channels should be used to apply stimulation.Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1

[0064] In one non-limiting example, the ENI-based channel selection module 262 can select the stimulation channels by examining the post-processed channelized signals 257 (or the filtered channelized signals 255), identifying the peaks in the frequency spectrum of the received sound signal, and selecting the stimulation channels corresponding to these peaks. In this example, the ENI-based channel selection module 262 selects the stimulation channel corresponding to the band-pass filtered signal with the largest amplitude. In other examples, the ENI-based channel selection module 262 can select the stimulation channels based on a current spread function (e.g., a spread-of-excitation (SOE) model). In some example embodiments, the ENI-based channel selection module 262 can comprise an ENI SNR estimator module 276, which is configured to estimate signal-to-noise ratios (SNRs) at the electrode-neural interface (ENI) based on the channel SNR values 259 generated and output by the channel SNR estimator 258. In such examples, the ENI-based channel selection module 262 can select the stimulation channels based on channel signal-to-noise ratio (SNR) values, cumulative SNR values at the electrode-neural interface (ENI SNR values), or combinations thereof, among various other factors / variables. The signals that are selected at the ENI-based channel selection module 262 are represented in FIG. 2A by arrow 263 and are referred to herein as selected channelized signals or, more simply, selected signals 263. However, it should be appreciated that different channel selection strategies can be used based on a SCAN output, including a conventional channel selection strategy and / or an ENI SNR based channel selection strategy. In noisy conditions, using the ENI SNR based channel selection strategy described herein can result in an improvement in “perceived SNR” at the neural elements, as compared to a channel selection strategy that selects n of m channels by selecting peaks, for example.

[0065] In the embodiment of FIG.2A, the ENI-based channel selection module 262 can select a subset ‘n’ of the ‘m’ post-processed channelized signals 257 for use in generation of electrical stimulation signals for delivery to a recipient (i.e., the sound processing channels are reduced from ‘m’ channels to ‘n’ channels). In one non-limiting illustrative example, the ‘n’ largest amplitude channels (maxima) from the ‘m’ available combined channel signals / masker signals is made, with ‘m’ and ‘n’ being programmable during initial fitting, and / or operation of the hearing device. In an example where ‘m’ is 22 channels, ‘n’ can correspond to 6 channels, 8 channels, 10 channels, 12 channel, etc. However, it is to be appreciated that channel selection is not limited to maxima selection techniques, and various other different channel selection methods could additionally or alternatively be used (e.g., based on a current spread functionAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 (SOE model), based on channel SNR values, based on ENI SNR values, etc.), as described herein.

[0066] The ENI-based channel selection module 262 of FIG. 2A, in combination with the channel SNR estimator 258, can be used to implement a dynamic “electrode-neural interface SNR” based sound coding strategy. In some non-limiting examples, the ENI-based channel selection module 262 can include or perform several functional operations (or sub-modules), including but not limited to, amplitude determination, current spread functions (spread of excitation (SOE) models), and maxima selection, in addition to the ENI SNR estimator module 276. In one example embodiment, the channel SNR estimator 258 estimates per-channel SNR values based on the post-processed channelized signals 257 (or the filtered channelized signals 255), and provides the channel SNR values 259 to the ENI SNR estimator 276. The ENI SNR estimator 276 receives output from the current spread function (SOE model), and also receives the channel SNRs 259 from the channel SNR estimator 258. The ENI SNR estimator 276 calculates the cumulative SNR at the ENI (also referred to as an ENI SNR value) based on the output from the current spread function module (SOE model) and the channel SNRs 259 received from the channel SNR estimator 258. The ENI SNR values can be used in an iterative process that selects the set of maxima (the appropriate channels / electrodes) for stimulation that optimizes / maximized the ENI SNR value. Thus, the ‘n’ maxima that are selected for stimulation according to this iterative process can be a particular combination of channels / electrodes (the selected signals 263) that will result in the best (largest / highest) SNR at the ENI, as opposed to only considering amplitudes alone.

[0067] The sound processing path 250 also comprises the mapping and encoding module 264. Although the mapping and encoding module 264 is shown as being integrated in the stimulation strategy adaptation module 260 in FIG.2A, it should be appreciated that example embodiments are not limited thereto, and that the mapping and encoding module 264 could be integrated in the ENI-based channel selection module 262 or could be a separate module. The mapping and encoding module 264 receives the selected signals 263 from the ENI-based channel selection module 262. The mapping and encoding module 264 is configured to map the amplitudes of the selected signals 263 (the subset of ‘n’ channels that the ENI-based channel selection module 262 selected from among the post-processed channelized signals 257 or the filtered channelized signals 255), which converts amplitudes in terms of decibels (dB) into current levels (weights) for the stimulation channels on which stimulation is to be applied, into a set of output control signals 265 (e.g., stimulation commands) that represent the attributes of the electricalAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 stimulation signals that are to be delivered to the recipient so as to evoke perception of at least a portion of the received sound signals. This channel mapping may include, for example, threshold and comfort level mapping, dynamic range adjustments (e.g., compression), volume adjustments, etc., and may encompass selection of various sequential and / or simultaneous stimulation strategies.

[0068] In the embodiment of FIG. 2A, the set of stimulation commands (control signals 265) that represent the electrical stimulation signals are encoded for transcutaneous transmission (e.g., via an RF link) to an implantable component 112 (refer to FIGs.1A-1D). This encoding is performed, in the specific example of FIG. 2A, at the mapping and encoding module 264. The selected stimulation channels and their corresponding determined current levels are encoded in the control signals 265 for transmission to the implantable component 112. As such, the mapping and encoding module 264 is sometimes referred to herein as a channel mapping and encoding module, and operates as an output block configured to convert the plurality of channelized signals (the selected signals 263) into the plurality of output control signals 265. The implantable component 112 then applies stimulation in accordance with the specified current levels via the corresponding electrode contacts indicated by the output control signals 265.

[0069] The stimulation strategy adaptation module 260, as enabled at least in part by the ENI- based channel selection module 262, is configured to adjust one or more operations performed in the sound processing path 250 so as to achieve the target stimulation strategy (i.e., adapt the stimulus resolution, the channel / electrode selection, the current level, the amplitude / magnitude, the pulse rate, the pulse width, the stimulation type, etc. of the electrical stimulation that is delivered to the recipient). In some examples, the stimulation strategy adaptation module 260 can adjust operations of the filter-bank module 254, the post-processing module 256, the channel SNR estimator 258, the ENI-based channel selection module 262, and / or the mapping and encoding module 264 to generate output signals representative of electrical stimulation signals according to the target stimulation strategy (that results in an optimized (maximized) signal-to-noise ratio at the electrode-neural interface).

[0070] The stimulation strategy adaptation module 260 may adjust operations of the sound processing path 250 at a number of different time scales. For example, the stimulation strategy adaptation module 260 may determine the target stimulation strategy and make corresponding processing adjustments in response to a triggering event, such as the detection of a change in the listening environment (e.g., when the channel SNR values, and / or optionally the soundAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 classification data, indicates the cochlear implant 102 is in a listening environment that is different from the previous listening environment). Alternatively, the stimulation strategy adaptation module 260 can determine the target stimulation strategy and make corresponding processing adjustments substantially continuously, periodically (e.g., every 1 second, every 5 seconds, etc.), according to a program or schedule, or the like, to adapt the sound processing operations accordingly (i.e., so that the resulting electrical stimulation signals delivered to the recipient correspond to the target stimulation strategy).

[0071] As noted, presented herein are techniques for maximizing the cumulative SNR at the electrode-neural interface (ENI) according to a maxima selection algorithm. As used herein, “ENI SNR” refers to the interference between multiple channels (e.g., overlapping current spread patterns, which can have a “blurring” or “smearing” effect with respect to adjacent channels). In some examples, this algorithm can involve estimating channel SNRs for each channel. When performing maxima selection based on amplitude alone (ACE) or based on current spread (SPACE / SOE) alone, signal-to-noise ratio (SNR) is not taken into account, such that the cumulative SNR at the ENI is not optimal in many instances. If a first channel has high SNR (good channel), and a second neighboring channel has low SNR (bad channel), then the cumulative SNR at the ENI is lower (neurons that are targeted see a worse SNR because of the combination of both the first / good channel and the second / bad channel targeting the same neurons).

[0072] One solution according to example embodiments described herein involves performing maxima selection based on channel SNR so as to optimize the cumulative SNR at the ENI. This can include selecting a different third channel with higher SNR (better channel) than the second neighboring channel (bad channel), and deselecting the second neighboring channel with low SNR (bad channel), so that the cumulative SNR at the ENI is higher. That is, the neurons that are targeted see a better SNR because of the combination of both the first / good channel and the third / better channel targeting the same neurons).

[0073] Another solution according to example embodiments described herein involves performing maxima selection based on current spread (SPACE / SOE) and channel SNR so as to optimize the cumulative SNR at the ENI. This can include selecting a different fourth channel with higher SNR (better channel) than the second neighboring channel (bad channel) and that has a greater separation / distance from the first channel with high SNR (good channel), and deselecting the second neighboring channel with low SNR (bad channel), so that the cumulative SNR at the ENI is higher. That is, the neurons that are targeted see a better SNRAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 because of the combination of both the first / good channel and the fourth / better / farther channel targeting the same neurons).

[0074] Next, an exemplary maxima selection process will be described with reference to the ENI-based channel selection module 262 of FIG.2A. (1) Obtain a first SNR for a first channel (various exemplary algorithms can be used to estimate noise or signal-to-noise ratio), and apply the first channel SNR to estimate the signal and noise components for the first channel (see Equation (d) in Appendix). Apply the current spread function (SOE model) to the signal and noise components of the first channel to calculate the ENI amplitudes (i.e., dampened signal and noise amplitudes) at other channel locations along the cochlea (see Equation (c) in Appendix). (2) Obtain a second SNR for a second neighboring channel, and apply the second channel SNR to estimate the signal and noise components for the second channel. Apply the current spread function (SOE model) to the signal and noise components of the second channel to calculate the ENI signal and noise amplitudes (dampened amplitudes). (3) Sum the first channel and the second channel ENI amplitudes (see Equation (e) in Appendix). Calculate a first cumulative SNR at the ENI (see Equation (f) in Appendix). Based on how the amplitude and channel SNR changes across the channels and the application of the current spread function, the cumulative SNR at the ENI will also change across the channels (see Equations (e) + (f) in Appendix) (4) Obtain a third SNR for a third channel, and apply a third channel SNR to estimate the signal and noise components for the third channel and calculate the ENI amplitudes for the third channel based on the current spread function (SOE model). (5) Sum the first channel and the third channel ENI amplitudes to calculate a second cumulative SNR at the ENI. (6) Repeat for additional channels (fourth, fifth, etc.), which could be some or all remaining channels. (7) Select the combination of channels that results in the highest cumulative SNR at the ENI (lowest noise / interference).

[0075] It should be appreciated that amplitudes of respective channels will make a difference in the cumulative SNR at the ENI (noisy channels with higher amplitudes will have more overlap / interference / blurring with neighboring channels, while noisy channels with lower amplitudes will have less overlap / interference / blurring with neighboring channels). Generally, the above-described technique involves estimating individual signal and noise amplitudes based on individual channels SNRs which are then used to calculate the damped values based on the individual current spread functions, obtaining a cumulative spread of excitation (SOE) pattern including all channels, and estimating cumulative SNR at ENI for that combination of channels. Based on this information, the algorithm can then choose which ‘n’ maximas (e.g., 8Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 channels, etc.) among the ‘m’ channel (e.g., 22 channels) are the best to combine, so that when all of those channels are superimposed together, an excitation pattern which has the highest cumulative SNR at the ENI (to reduce / limit / minimize interference, noise, overlap, etc.) is obtained as a result of the maxima selection algorithm implemented by the ENI-based channel selection module 262. It is noted that the term “excitation pattern” can refer to the “ENI signal” as used herein (i.e., the dampened values after applying the SOE model), and that these terms have similar meaning and can be used interchangeably in the present disclosure.

[0076] As described elsewhere herein, the stimulation strategy adaptation module 260 may set or adjust various operations of the sound processing path 250, such as the operations of the filter-bank module 254, the post-processing module 256, the ENI-based channel selection module 262, and / or the mapping and encoding module 264, to set the stimulation strategy (e.g., the stimulus resolution, the channel selection, etc.) of the delivered electrical stimulation signals. In one embodiment, the spatial / spectral attributes of the stimulus resolution are set by switching between different channel / electrode configurations, such as between monopolar stimulation, wide / defocused stimulation, focused (e.g., multipolar current focusing) stimulation, etc.

[0077] FIG. 2B is a flowchart illustrating a method 290 in accordance with embodiments presented herein. Method 290 begins at operation 292, where a hearing device located in an acoustic environment receives input sound signals for analysis. At operation 294, the hearing device assesses the acoustic environment based on the input sound signals, and determines (calculates / estimates) channel SNR values associated with the input sound signals. Optionally the hearing device can also determine a sound class of the input sound signals (e.g., speech, speech-in-noise, noise, music, quiet). At 296, a target stimulation strategy (e.g., stimulus resolution, channel selection, etc.) for the generation of electrical stimulation signals is set or adjusted based on the assessment of the acoustic environment (i.e., the channel SNR values, the sound class of the input sound signals, ENI SNR values, etc.). The target stimulation strategy can correspond to a particular subset of channels / electrodes that will optimize (maximize) SNR at the electrode-neural interface (ENI), based at least in part on the channel SNR values. The hearing device is configured to adapt the sound processing path differently depending on the determined channel SNR values (and optionally, the determined sound class). For example, for a speech sound class or a speech in noise sound class, channel SNR values of the input sound signals can be estimated and then used to determine or control the target stimulation strategy (e.g., set / adapt stimulus resolution, optimize channel selection, etc.). TheAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 channel SNR values provide a measure of how much speech compared to noise is present in the input sound signals. At operation 298, the hearing device generates electrical stimulation signals that are representative of the input sound signals according to the target stimulation strategy. The electrical stimulation signals are then delivered to a recipient of the hearing device. Thus, the hearing device can select an optimal configuration of channels / electrodes based on the SNR values (e.g., the ENI SNR values that are based on the channel SNR values) of the input sound signals.

[0078] The method 290 may be implemented using a hearing device that comprises: one or more sound input elements configured to receive sound signals; a sound processing path configured to convert the sound signals into one or more output signals for use in delivering electrical stimulation to a recipient; and a stimulation strategy adaption module configured to set or adjust a target stimulation strategy of the electrical stimulation based on the SNR values of the sound signals. In some examples, a channel SNR estimator module is configured to determine a presence of speech, noise, or a combination of speech and noise, classify the sound signals as speech signals, noise signals, or a combination of speech and noise signals, and determine signal to noise ratios of the signals. In such examples, the stimulation strategy adaption module is further configured to select the electrode configuration for use in delivering the electrical stimulation to the recipient via the one or more stimulation channels based on the SNR values of the signals.

[0079] As described in detail above, presented herein are techniques that analyze the acoustic scene / environment of a hearing device and, accordingly, adjust, adapt, or otherwise set the electrical stimulation strategy based on the acoustic environment (e.g., adjust stimulus resolution, channel selection, etc. based on an estimated listening difficulty that the acoustic environment presents to a recipient of the hearing prosthesis). The techniques presented herein leverage the idea that, in more challenging listening situations that are more difficult for recipients, it may be beneficial to use more power in order to create a more accurate neural activation pattern that lessens the listening burden on the recipient. Accordingly, the techniques presented optimize power consumption and hearing performance based on the listening situation.

[0080] As described further below, FIGs. 3A, 3B, 4A, 4B, 5A and 5B generally depict exemplary electrode currents and stimulation patterns for different channel configurations, in which a plurality of electrodes spaced along the recipient’s cochlea frequency axis (i.e., along the basilar membrane), and include solid lines of varying lengths that extend from variousAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 electrodes to illustrate the intracochlear stimulation current delivered in accordance with a particular channel configuration. However, it is to be appreciated that the stimulation patterns shown are generally illustrative in nature and that, in practice, the stimulation current can spread differently in different recipients and according to different channel selections, stimulation types, physiological factors, etc. It is also to be appreciated that stimulation can be delivered to a recipient using charge-balanced waveforms, such as biphasic current pulses and that the length of the solid lines extending from the electrodes illustrate the relative “weights” that are applied to both phases of the charge-balanced waveform at the corresponding electrode in accordance with different channel configurations. As described further below, the different stimulation currents (i.e., different channel weightings) result in different stimulation patterns, respectively, of voltage and neural excitation along the frequency axis of the cochlea.

[0081] For example, with the use of a monopolar channel configuration, all of the intra- cochlear stimulation current is delivered with the same polarity via a single electrode, and generates a stimulation pattern which spreads across neighboring electrodes. The stimulation pattern represents the spatial attributes (spatial resolution) of the monopolar channel configuration. In wide or defocused channel configurations, the stimulation current is split amongst an increasing number of intracochlear electrodes, and the width of the stimulation patterns increases, thus providing increasingly lower spatial resolutions. In general, the wider the stimulation pattern, the lower the spatial resolution of the stimulation signals. With focused channel configurations, intracochlear compensation currents are added to decrease the spread of current along the frequency axis of the cochlea, where the compensation currents are delivered with a polarity that is opposite to that of a primary / main current. In general, the more compensation current at nearby electrodes, the more focused the resulting stimulation pattern (i.e., the spatial resolution is increased by introducing increasingly large compensation currents on electrodes surrounding the central electrode with the positive current. Thus, the intra- cochlear stimulation current generates a stimulation pattern that is generally localized to the spatial area adjacent the electrode to which positive stimulation current is delivered.

[0082] As noted, presented herein are techniques for selecting stimulation channels for application of stimulation by a hearing device or other medical device, such as a cochlear implant or an auditory brain stimulator. The hearing device filters a received signal to obtain a plurality of band-pass filtered signals, each corresponding to one or more stimulation channels. The hearing device then selects a stimulation channel for application of stimulation based on various factors, including but not limited to, amplitudes, current spread functionsAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 (spread-of-excitation (SOE) models), signal-to-noise ratio (SNR), electrode-neural interface (ENI) factors, etc. In some example embodiments, the ENI factors can include a cumulative SNR at the ENI (also referred to herein as an “ENI SNR” value).

[0083] Example embodiments involve delivering stimulation with regard to calculating an effective SNR at the neural output (as opposed to merely channel output). The present disclosure focuses on making adjustments to the channel selection. Other adjustments are also possible, including channel amplitudes (current levels), stimulus resolution, pulse rate, pulse width, etc. In some example embodiments, machine learning technologies (e.g., neural networks, etc.) can be implemented, where the cost function is the SNR at the ENI, for example.

[0084] According to one example embodiment, a technique is provided to improve effective signal-to-noise ratio (SNR) at the electrode-neural interface (ENI). FIG. 3A illustrates an example in which channels are selected and stimulated based on their amplitudes. The size / length of the arrows correspond to channel amplitude (e.g., a1 denotes a smaller amplitude, while a2 denotes a larger amplitude). In this example, channel selection is performed based on the amplitudes 345(2) and 345(3) of the channels 344(2) and 344(3), respectively. FIG.3A is an example in which there is lower SNR at the electrode-neural interface (ENI), as represented by the signal 349(A) (clean+noisy) which results from the substantial overlap of signal 347(3) (noisy) with signal 347(2) (clean). FIG. 3B illustrates an example in which channels are selected and stimulated based on signal-to-noise ratios (SNRs), according to an example embodiment. This is similar to speech in quiet stimulation, by comparison. In this example, channel selection is performed based on the local channel SNR values of the channels 344(2) and 344(1), respectively, as opposed to their amplitudes 345(2) and 345(1) alone. In this instance, channel 344(1) is selected for stimulation instead of 344(3), since channel 344(1) has a higher channel SNR value, even though channel 344(1) has a smaller amplitude. FIG.3B is an example in which there is better SNR at the electrode-neural interface (ENI), as represented by the signal 349(B) (clean+clean) which results from the overlap of signal 347(1) (clean) with signal 347(2) (clean).

[0085] As shown in FIG. 3B, performing channel selection for stimulation based on the local channel “signal-to-noise ratio” (rather than based on the “amplitude” of the channel alone, as shown in FIG. 3A) will prevent the noise dominant channels from interfering with the information in the signal dominant channels (or at least reduce, limit, or minimize such interference), thereby improving the overall SNR observed at the electrode-neural interface (ENI). Thus, the effective SNR at the electrode-neural interface (ENI) can be improved byAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 performing channel selection based on the local channel SNR (and not only based on the amplitude of the channel), such that channels with better SNR are selected for stimulation (rather than selecting channels with the largest / highest amplitude), according to the example embodiment of FIG. 3B. Improved SNR estimation techniques can also be used to reduce / limit / minimize errors when estimating the local channel SNR values.

[0086] According to another example embodiment, a technique is provided to improve channel independence based on signal-to-noise ratio (SNR). FIG. 4A illustrates an example in which channels 444(2) and 444(3) are selected and stimulated based on their amplitudes 445(2) and 445(3). As shown in FIG. 4A, there is relatively high blurring of the two signals 447(2) and 447(3), due to the relatively short separation distance (d1) between channel 444(2) and channel 444(3), as represented by the signal 449(A) which results from the substantial overlap of signal 447(3) with signal 447(2). FIG.4B illustrates an example in which channels 444(2) and 444(4) are selected and stimulated based on channel independence, according to an example embodiment. As shown in FIG. 4B, there is relatively low blurring of the two signals 447(2) and 447(4), due to the relatively longer separation distance (d2) between channel 444(2) and channel 444(4), as represented by the signal 449(B) which results from the minimal overlap of signal 447(4) with signal 447(2).

[0087] As shown in FIG. 4B, increasing the electrode separation between the actively stimulating electrodes will improve the channel independence factor by reducing the overlap of current spread patterns (lower / reduced blurring effect) at the electrode-neural interface (ENI). For “speech-in-quiet” conditions, the example technique of FIG.4B improves channel independence only. This technique could be used in conjunction with, or as an alternative to, the use of current focusing techniques. Thus, the channel independence factor can be improved (for speech-in-quiet conditions) by increasing the distance between the active stimulating electrodes (i.e., increased electrode separation between stimulated channels), according to the example embodiment of FIG.4B.

[0088] FIG. 5A illustrates an example in which channels 544(2) and 544(3) are selected and stimulated based on their amplitudes 545(2) and 545(3). FIG.5A is an example in which there is low channel independence, and lower SNR due to interference from the low SNR channel, as represented by signal 549(a) (clean+noisy) which results from the substantial overlap of signal 547(3) (noisy) with signal 547(2) (clean). As shown in FIG.5A, in a noisy background, the information (clean signal) is blurred not only by the signal from the neighboring channels, but is also blurred by the noise from the neighboring channels, due to the relatively shortAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 distance (d1) between channel 544(2) with the clean signal 547(2) and channel 544(3) with the noisy signal 547(3). FIG. 5B illustrates an example in which channels 544(2) and 544(4) are selected and stimulated based on signal-to-noise ratio (SNR) and channel independence, according to an example embodiment. FIG.5B is an example in which there is higher channel independence, and higher SNR due to less interference from the low SNR channel, as represented by signal 549(B) (clean+noisy) which results from the minimal overlap of signal 547(4) (noisy) with signal 547(2) (clean). As shown in FIG.5B, the information (clean signal) is less blurred by the noisy signal from the neighboring channel, due to the relatively longer distance (d2) between channel 544(2) with the clean signal 547(2) and channel 544(4) with the noisy signal 547(4).

[0089] As shown in FIG. 5B, increasing the electrode separation between the actively stimulating electrodes will prevent the noise dominant channels from interfering with the information in the signal dominant channels (or at least reduce, limit, or minimize such interference), thereby improving the overall SNR observed at the electrode-neural interface (ENI). For “speech-in-noise” conditions, the example technique of FIG.5B will improve both the channel independence and the effective SNR at the electrode-neural interface (ENI). Thus, the effective SNR (of the blurred representation) at the electrode-neural interface (ENI) can also be improved to some extent (for speech-in-noise conditions) by increasing the distance between the active stimulating electrodes (i.e., increased electrode separation between stimulated channels), according to the example embodiment of FIG.5B.

[0090] One or more of the concepts described above with reference to FIGs. 3B, 4B, and / or 5B can be combined, as further described below with reference to the example embodiment of FIG. 6A. FIG. 6A is a block diagram of a stimulation strategy adaptation module 660 for implementing a greedy search maxima selection algorithm, according to an example embodiment, which is configured to search for ‘n’ maxima out of ‘m’ total channels. The objective of the greedy search maxima selection algorithm could be to either: (a) maximize effective SNR at the ENI, (b) maximize channel independence, or (c) strike an appropriate tradeoff / balance between the effective SNR at the ENI and channel independence based on a metric.

[0091] The stimulation strategy adaptation module 660 of FIG.6A includes a channel signal- to-noise ratio estimator module 658 (also referred to as channel SNR estimator 658), and an ENI-based channel selection module 662. In some example embodiments, operations and resulting outputs of the greedy search maxima selection algorithm implemented by the ENI-Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 based channel selection module 662 can be a machine-learned stimulation logic, and / or can be used to train a machine learning stimulation device, as described elsewhere herein.

[0092] In the example of FIG. 6A, a filter-bank (not shown, refer to FIG. 2A) provides input (e.g., channelized signals 657) to the channel SNR estimator 658 and the ENI-based channel selection module 662. In this example, the ENI-based channel selection module 662 includes a current spread function 674 (SOE model) and an electrode-neural interface signal-to-noise ratio estimator module 676 (also referred to as ENI SNR estimator 676).

[0093] The current spread function 674 (SOE model) can enable individualized maxima selection 681, and can also provide a channel independence measure 683 (e.g., via electrode voltage telemetry (EVT), electric field imaging, trans-impedance matrix (TIM) measurement, etc.). The output 675 of the current spread function 674 (e.g., individualized maxima selection 681 and / or channel independence measure 683) is provided as input to the ENI SNR estimator 676.

[0094] The channel SNR estimator 658 estimates (calculates) a channel signal-to-noise ratio (SNR) for the channels / electrodes. The output 659 of the channel SNR estimator 658 (e.g., the channel SNRs for each of the channels / electrodes) is provided as input to the ENI SNR estimator 676. In some example embodiments, the output 659 (channel SNRs) could also be provided to the current spread function 674 (SOE model) by the channel SNR estimator 658, so that the current spread function 674 can dynamically apply the channel SNRs (e.g., when modeling the spread of excitation (SOE) patterns of the signals).

[0095] The electrode-neural interface SNR estimator 676 estimates (calculates) an effective signal-to-noise ratio (SNR) at the electrode-neural interface (ENI), based on the inputs received from the current spread function 674 and the channel SNR estimator 658, respectively. Thus, the ENI SNR estimator 676 of the stimulation strategy adaptation module 660 can implement a greedy search maxima selection algorithm such that the output 663 improves (or can be used to maximize) the “effective” SNR at the ENI 687, also referred to herein as an “overall” SNR at the ENI, a “cumulative” SNR at the SNI, or more simply, an “ENI SNR” value.

[0096] Next, an exemplary iterative process of the ENI SNR estimator 676 is described with reference to the example of FIG. 6A: (1) start with selecting a first channel with the highest channel SNR as a maxima for inclusion in a final set of channels (n of m); (2) then add one additional channel at a time, and estimate the ENI SNR; repeat for all channels; (3) select a second channel with the second highest ENI SNR as a maxima for inclusion in the final set ofAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 channels (n of m); (4) with the first and second channels selected, add one additional channel at a time and estimate the ENI SNR; repeat for all channels; (5) select a third channel with the third highest ENI SNR for inclusion as a maxima in the final set of channels (n of m); and (6) repeat until n channels (with highest ENI SNR) are selected to obtain the final set of channels for stimulation. With regard to channel independence, channel selection could be based on regions of array (apical, middle, basal), or minimum threshold separation / distance (e.g., at least 2, 3, 4 electrodes apart, etc.), as described.

[0097] There are also other ways to approach the channel selection problem by incorporating SNR measures. For example, another exemplary iterative process can include: (1) select maxima based on amplitude (initial set of channels); (2) adjust maxima selection based on channel independence (e.g., such that no immediately neighboring channels are selected), such as by selecting a non-adjacent channel (possibly with lower amplitude) instead of an adjacent channel (with higher amplitude); (3) adjust maxima selection based on cumulative SNR at the ENI (e.g., such that a high amplitude, low channel SNR is deselected, and / or such that a low amplitude, high channel SNR is selected instead), such as by selecting a channel with higher SNR (less noisy, but possibly with lower amplitude) instead of a channel with lower SNR (higher amplitude, but more noisy); and (4) adjust maxima selection by balancing both channel independence and cumulative SNR at the ENI according to a metric. For example, the system should avoid selecting neighboring channels as much as possible (separation / distance of at least 1-2 electrode(s)), with an exception that a neighboring / adjacent channel with high channel SNR can be selected as maxima.

[0098] FIG. 6B is a flowchart of an example method 690 for enhanced speech in noise perception, according to an example embodiment. Method 690 begins at operation 691, where channelized signals are determined from an input signal. At operation 693, channel signal-to- noise ratios (SNRs) are estimated for a plurality of the channelized signals. At operation 695, current spread functions are determined for at least a subset of the plurality of channelized signals. At operation 697, electrode-neural interface (ENI) signal-to-noise ratios (SNRs) are estimated for at least the subset of the plurality of channelized signals based on the current spread functions and the channel SNRs. At operation 699, a set of the channelized signals are selected, based on the ENI SNRs, for use in delivering stimulation signals to a recipient.

[0099] In some examples, once the SNRs are calculated, the signal and noise components of the channel amplitudes are estimated at operation 693. In some examples, current spread functions could be either fixed or customized based on subjects, and / or could be determinedAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 based on EVT, TIM, etc., at operation 695. In some examples, there can also be an implicit operation 696 (not shown in FIG. 6B) that occurs between operation 695 and operation 697 and involves computing the excitation patterns (or ENI amplitudes) of signal / noise, for use in estimating the ENI SNRs for the channelized signals.

[0100] According to another example embodiment, a technique is provided for spread pre- compensation advanced combination encoder (SPACE) maxima selection by incorporating signal-to-noise ratio (SNR) measures, which is sometimes referred to herein as an “SNR- SPACE” sound coding strategy. As shown in the graph 700 of FIG. 7A, a spatial advanced combination encoder (SPACE) algorithm assumes a standard spread of excitation (SOE) 710 for each channel by default (represented by curve 710).

[0101] Also shown in the graph 700 of FIG. 7A, a modified SPACE algorithm (“SNR- SPACE”) according to an example embodiment assumes a greater spread of excitation (SOE) 720 for channels with high SNR (represented by curve 720). To implement this, channels with high channel SNRs are “forced” to assume a larger current spread (SOE). This greater SOE assumed by the model indirectly assigns a greater weight to the “high SNR” channels in the current spread function (SOE model) calculation, which effectively increases the chances of the “high SNR” channels getting selected by the greedy search maxima selection algorithm for the maxima selection. Likewise, some example embodiments can “force” channels with low channel SNRs to assume a smaller current spread (SOE), which will indirectly assign a lesser weight to the “low SNR” channels in the current spread function (SOE model) calculation, and thereby effectively decreases the chances of the “low SNR” channels being selected by the greedy search maxima selection algorithm.

[0102] In one possible alternative example embodiment, high SNR channels could be forced to have a higher amplitude, and thereby increase the chances of the high SNR channels being selected. Likewise, low SNR channels could be forced to have a lower amplitude, and thereby decrease the chances of the low SNR channels being selected. Relatively minor increases / decreases of amplitude can be used (e.g., 10% attenuation / change), for example.

[0103] FIG. 7B is an example of a signal-to-noise ratio (SNR) based SPACE sound coding strategy (SNR-SPACE), which performs maxima selection by incorporating SNR measures, in accordance with an example embodiment. FIG.7B illustrates stimulation levels (vertical lines with symbols at the endpoints) and excitation patterns (curved lines) as a function of the channels (numbered electrodes 1-22). The solid curved line represents the original envelopeAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 input spectrum 741, while the height of each vertical line ending at the symbols indicates the stimulation levels (amplitudes) as a percentage of the recipient’s electrical dynamic range (in % DR) between threshold (T) level and comfort (C) level. The dashed curved line represents the modeled excitation pattern 743 computed with the corresponding fraction of current spread (using SNR-SPACE).

[0104] With a standard clinical advanced combination encoder (ACE) strategy, channels are selected based on amplitude only, independent of any SNR measures. In ACE, maxima selection retains a subset of ‘n’ channels (out of ‘m’ total available channels) corresponding to the spectral maxima (preserving their amplitude). The ACE strategy would select the spectral maxima (e.g., channels 4-11 in FIG. 7B), regardless of their spatial distribution and / or their SNR. However, channel selection based on the spectral maxima alone may not be the most optimal features for speech perception in many instances. Since the ‘n’ channels with the highest short-term envelope magnitudes are selected irrespective of their spatial distribution, they tend to be clustered, this can lead to a more pronounced channel interaction in the modeled excitation pattern. Due to the effect of the spread, the excitation pattern resembles a spatially blurred version of the stimulation pattern (where the degree of blurring increases as the spread gets wider / broader). Spectral blurring is caused by the limitations at the electrode-neural interface (ENI), such as a large spread of excitation (SOE).

[0105] Instead of channel selection based on spectral maxima, a spatial advanced combination encoder (SPACE) strategy integrates a spread pre-compensation algorithm into the ACE algorithm. With SPACE, channels are selected based on a spread of excitation (SOE) model, independent of any SNR measures. In SPACE, amplitude-based maxima selection can be replaced by a spread pre-compensation channel selection (SPCS) algorithm, for example, in which a model of the current spread is used to estimate the appropriate selection of channels and associated stimulation levels such that the resulting excitation pattern approximates the input envelope spectrum. That is, the SPCS algorithm pre-compensates the stimulation pattern for the spectral blurring caused by the SOE to match the excitation pattern of the enhanced signal to the original (unblurred) spectrum. In SPACE, the channel selection is performed to retain only the subset ‘n’ of ‘m’ channels with pre-compensated stimulation level that are greater than or equal to the threshold (T) level. In a high spread pre-compensation program (H-SPACE), fewer channels are selected as the degree of pre-compensation increases. This results in a selection of channels that are more widely distributed across the array and where local spectral peaks are generally preserved, while valleys are either attenuated or not selectedAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 for stimulation. Thus, the SPACE strategy “sharpens” the stimulation pattern such that the resulting excitation pattern approximates the input spectrum. The SOE determines not only which channels and the number of channels that are selected, but also the adjustment of the associated stimulation levels. In this way, a finer control over the overall excitation pattern that is produced can be achieved, analogous to the application of a soft gain (with some value between 0 and 1) to each channel (as opposed to a binary decision).

[0106] According to the SNR-SPACE sound coding strategy in the example of FIG. 7B, channels can be selected based on a combination of channel amplitudes, current spread functions (spread of excitation (SOE) models), channel signal-to-noise ratio (SNR) measures, and / or electrode-neural interface (ENI) SNR measures. In this manner, the example SNR- SPACE sound coding strategy of FIG. 7B is dependent on the signal-to-noise ratios of the channels / electrodes, in contrast with the standard ACE strategy and the SPACE strategy (which do not factor in SNR values).

[0107] At the upper left of the graph in FIG. 7B, the modeled excitation pattern shows the effect 749 of artificially increasing the SOE on a high SNR channel (e.g., channel 744(6) in the example of FIG. 7B). In this example, the low SNR channel 744(5) that would otherwise be selected using SPACE is not selected or is deselected using SNR-SPACE, and the high SNR channel 744(6) is selected instead. In addition, channel 744(15) and channel 744(21) are also selected using SNR-SPACE based on analysis of the current spread function. Thus, the spread of excitation (SOE) assumed for each channel by using the modified SPACE strategy (“SNR- SPACE”) of FIGs. 7A and 7B can be adaptively controlled (e.g., artificially increased or decreased) based on channel SNR (in addition to control based channel amplitude and the current spread function / SOE model), such that channels with higher SNR are assumed to have a greater SOE, and hence are weighted higher for maxima selection. As noted, improved SNR estimation techniques can also be used to reduce / limit / minimize errors when estimating the channel SNR values.

[0108] As noted, presented herein are techniques for use of machine learning for improving speech in noise perception by incorporating electrode neural interface factors for stimulation strategies, including for detection of events and adaption of operation of an implantable medical device system according to the detected events. Examples of events can include, but are not limited to, the presence of speech in various noisy conditions. As noted, aspects of the techniques presented herein use machine learning to automatically detect an event, such as a speech-in-noise event (e.g., detecting the presence of both speech and background noise in aAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 sound environment). Once a particular event is detected, the techniques presented herein can adjust operation of the cochlear implant, hearing device, or medical device to deliver a treatment / therapy to the user (e.g., deliver a stimulation therapy), where the attributes of the delivered therapy are selected (adjusted) based on attributes of the detected event, and (optionally) user preferences. Stated differently, the machine learning techniques presented herein allow for the selection of a therapy that is optimized for the specific detected event and / or for the specific user (e.g., to account for the user’s stimulation preferences).

[0109] FIG. 8A illustrates a specific use of the techniques presented to select an optimal stimulation strategy for a user. That is, FIG. 8A is a functional block diagram illustrating an example stimulation system 802 configured with machine-learning stimulation logic, such as machine-learning stimulation module 860, for automated selection of stimulation strategies in response to detected events (e.g., various speech-in-noise conditions). The stimulation system 802 could be a stand-alone implantable stimulation device, or incorporated as part of an auditory prosthesis, such as a cochlear implant, bone conduction device, middle ear auditory prosthesis, direct acoustic stimulator, auditory brain stimulator, etc.

[0110] It is to be appreciated that the functional blocks illustrated in FIG. 8A can be implemented across one or more different devices or components that can be implanted in, or external to, the body of a user. The stimulation system 802 can comprise or be a component of, for example, a medical device system (e.g., a cochlear implant system), a computing device, a consumer electronic device, etc. As shown, the stimulation system 802 comprises a sensor unit 864, a processing unit 866, and a stimulation unit 868. Again, the sensor unit 864, the processing unit 866, and the stimulation unit 868 can each be implemented across one or more different devices and, as such, the specific configuration shown in FIG. 8A is merely illustrative.

[0111] The sensor unit 864 comprises a plurality of sensors 865(1)-865(N) that are each configured to capture signals representing an ambient / external sound environment of the user. The signals captured by the sensors 865(1)-865(N) are “state data” or “state variables” 879 (refer to FIG. 8B) and can take a number of different forms and can be captured by a number of different sensors. For example, the sensors 865(1)-865(N) can comprise sound sensors (e.g., microphones capturing sound signals), movement sensors (e.g., accelerometers capturing accelerometer signals), body noise sensors, medical sensors, such as electroencephalogram (EEG) sensors (e.g., one or more external or implantable electrodes and one or more associated recording amplifiers configured to record / measure electrical activity in the user’s brain),Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 electromyography (EMG) sensors or other muscle or eye movement detector (e.g., one or more external or implantable electrodes and one or more associated recording amplifiers configured to record / measure muscle response or electrical activity in response to a nerve's stimulation of the muscle), photoplethysmography (PPG) sensor (e.g., sensors configured to optically detect volumetric changes in blood in peripheral circulation), electro-oculogram (EOG) sensors, polysomnographic sensors, magnetoencephalography (MEG) sensors, heart rate sensors, temperature sensors, skin conductance sensors, functional near-infrared spectroscopy (fNIRS) sensors, etc. (e.g., recording heart rate, blood pressure, temperature, etc.). It is to be appreciated that this list of sensors is merely illustrative and that other sensors can be used in alternative embodiments.

[0112] It is to be appreciated that the state data / variables 879 can also include not only the direct sensor signals, but also processed version of the sensor signals. For example, in certain embodiments, the state data / variables 879 can include sound / environmental classification data generated from captured sound signals. In these embodiments, a sound classification module is configured to evaluate / analyze the sound signals and determine the sound class of the sound signals. That is, the sound classification module is configured to use the received sound signals to “classify” the ambient sound environment and / or the sound signals into one or more sound categories (i.e., determine the input signal type). The sound classes / categories may include, but are not limited to, “Speech,” “Speech+Noise,” “Noise,” “Music,” and “Quiet.” The sound classification module can also estimate the signal-to-noise ratio (SNR) of the sound signals. According to some example embodiments, the sound classification module can be or include a channel SNR estimator configured to estimate a per-channel SNR of channelized sound signals, also referred to elsewhere herein as a “channel SNR”), and / or can be or include an ENI SNR estimator configured to estimate a cumulative SNR at the electrode-neural interface (ENI), when generating sound classification data that can be part of the state data / variables 879. However, it should be appreciated that classifying / categorizing sounds is not a requirement in the example embodiments and corresponding techniques described herein, and that the techniques involving SNR estimation (channel SNR values and / or ENI SNR values) can be performed without the need for any analysis for determining sound classes / categories.

[0113] In FIG. 8A, the state data / variables 879 captured by, or generated from, the sensors 865(1)-865(N) are converted into electrical input signals (if not already in an electrical form). As shown, the state data / variables 879 (electrical input signals) is / are provided to the machine- learning stimulation device 862. As shown in FIG.8A, the processing unit 866 comprises theAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 machine-learning stimulation module 860, a control module 872, and a remote control module 878. It is to be appreciated that the functional arrangement shown in FIG. 8A is merely illustrative and does not require or imply any specific structural arrangements. The various functional modules shown in FIG. 8A can be implemented in any combination of hardware, software, firmware, etc., and one or more of the modules could be omitted in different embodiments.

[0114] FIG. 8B is a functional block diagram illustrating training and final operation of the machine-learning stimulation module 860 of FIG. 8A, in accordance with embodiments presented herein. More specifically, the machine-learning stimulation module 860 shown in FIG.8B includes a state observing unit 882, a label data unit 884, and a learning unit 886. The machine-learning stimulation module 860 is configured to generate device configuration data 869 (e.g., one or more control outputs) representing at least a selected treatment / therapy for use by the system (implantable medical device) to improve perception of speech in noise by the user. Stated differently, the machine-learning stimulation module 860 is configured to determine an optimal stimulation strategy for use by the system to stimulate the user.

[0115] In the example of FIG.8B, the learning unit 886 receives inputs from the state observing unit 882 and the label data unit 884 in order to learn to detect an event, such as a speech-in- noise event, and to set / determine a stimulation strategy that is delivered to the user in response to the detected event, and that accounts for attributes of the event. In particular, the state observing unit 882 provides state data / variables 879 to the learning unit 886. The state data / variables 879 includes event data, which can include environmental data representing the current ambient environment of the user, such as the current external sound environment of the user, etc. The learning unit 886 can also receive operating state data 877 representing a current operating state of the system (e.g., stimulation system / apparatus), such as signal-to-noise ratio (SNR) values, and uses the operating state data 877 to set a stimulation strategy that is delivered to the recipient.

[0116] Through machine-learning techniques, the learning unit 886 correlates the state data / variables 879 and the label data 885, over time, to develop the ability to automatically detect the occurrence of a specific event and to automatically select an optimal stimulation strategy for the user, given the specific attributes of the detected event (e.g., a particular speech- in-noise condition). As described above, the learning unit 886 generates the device configuration data 869 from the state data / variables 879, the label data 885, and, in certain examples, the operating state data 877 (e.g., channel SNR values, ENI SNR values, etc.). AlsoAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 as noted above, the system is trained to automatically identify the event (e.g., a speech-in-noise event), and select the optimal stimulation strategy to address that event based on the training data (e.g., to improve perception of speech in noise according to electrode-neural interface factors).

[0117] Referring to FIGs. 8A and 8B, the machine-learning stimulation module 860 (via learning unit 886) uses the state data / variables 879, the label data 885, and potentially the operating state data 877, to determine whether an event (e.g., a speech-in-noise condition) is present, and to generate device configuration data 869 based on this determination, which is used to generate stimulation signals 883 for delivery to the user. That is, as noted, the device configuration data 869 represents (or approximates) the optimal stimulation settings / program, as determined through a machine-learning process, such as the one described with reference to FIG.8B.

[0118] Referring again to FIG. 8A, the control module 872 is configured to use the device configuration data 869 to select, set, determine, or otherwise adjust a stimulation strategy for the user, as a function of the detected event (e.g., a speech-in-noise condition), as determined by the machine-learning stimulation module 860. Stated differently, the stimulation signals that are to be provided to the user are specifically determined and adjusted or adapted, in real- time, based on the ambient / external sound environment of the user (including SNR values, for example), as determined by the machine-learning stimulation module 860.

[0119] In accordance with embodiments presented herein, the optimal stimulation strategy includes the delivery of stimulation signals 883 to the user. These stimulation signals 883 are generated by the stimulation unit 868. The stimulation signals can have a number of different forms (e.g., electrical stimulation signals, electro-mechanical stimulation signals, electro- acoustic stimulation signals, etc.), and underlying objectives. For example, in certain embodiments, the stimulation signals 883 can improve the user’s perception of speech despite the presence of noisy background conditions.

[0120] In the example of FIG.8A, the stimulation system 802 includes the stimulation unit 868 that is configured to generate the stimulation signals 883. The stimulation unit 868 operates based on control signals 881 from the control module 872. The control signals 881 can dictate a number of different attributes / parameters for the stimulation signals 883, and can also set modulations in the stimulation signals 883, transitions, etc.Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1

[0121] As noted, the machine-learning stimulation module 860 is trained to determine electrode-neural interface factors (e.g., estimate SNR at the ENI based on channel SNRs, and possibly in combination with current spread functions (SOE models) in some embodiments) and to determine the optimal stimulation strategy for the user (e.g., implement channel selection to improve perception of speech in noise based on channel SNR values and / or ENI SNR values). In certain embodiments, in addition to channel selection for stimulation, the machine-learning stimulation module 860 can be trained to dynamically adjust a current level (amplitude), a stimulus resolution, a frequency or modulation, a current level, pulse rate or pulse width, a stimulation type (e.g., monopolar or focused), etc. with respect to the stimulation signals 883.

[0122] In the specific example of FIG. 8A, the control module 872 is configured to store a plurality of different stimulation maps 875. In general, each of the stimulation maps 875 is a set / collection of parameters that, when selected, are used to generate the control signals 881, which control the generation of the stimulation signals 883 by the stimulation unit 868. The parameters can control the sound type, fluctuation or modulation rate, amplitude, sound level settings, on / off, pitch settings, transition time settings, etc. In operation, different stimulation maps 875 can be created (e.g., by the software, an audiologist / clinician, through artificial intelligence, etc.) for different situations (i.e., different combinations of ambient / external sound environmental classifications).

[0123] In the example of FIG.8A, the machine-learning stimulation module 860 can be trained to select one of the stimulation maps 875 for use in generating the stimulation signals 883 delivered to the user and / or dynamically adjust settings / attributes of the stimulation signals 883. However, it is to be appreciated that the presence of multiple stimulation maps is merely illustrative and that in some other embodiments, the machine-learning stimulation device 862 is trained to dynamically determine the settings / attributes for control signals 881 that are used to generate the stimulation signals 883 and / or dynamically adjust settings / attributes of the stimulation signals 883, without the use of stored stimulation maps.

[0124] In certain examples, selected stimulation parameters / settings can be used to provide stimulation until the device configuration data 869 from the machine-learning stimulation module 860 changes in manner that causes the control module 872 to select or adjust the stimulation strategy. Once the stimulation adjustment is selected for use, the control module 872 could manage the transition between the parameters / settings to avoid unintended issues (e.g., annoyance to the user, painful stimulation, abrupt transitions, etc.).Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1

[0125] As noted, the processing unit 866 of FIG. 8A also comprises a remote control module 878. In certain embodiments, the remote control module 878 can be used to update / adjust, over time, which stimulation map is selected by the control module 872 based, for example, on user preferences. That is, the remote control module 878 can be used as part of the training process described with reference to FIG. 8B to, for example, receive control data from an external device (e.g., mobile phone) operating with the stimulation system 802.

[0126] In certain examples, the stimulation system 802 is configured to deliver stimulation signals to the user in order to improve perception of speech in noise by the user, by incorporating electrode-neural interface factors. While the stimulation signals are delivered to the user, one or more attributes / parameters of the stimulation signals (e.g., amplitude, resolution, channel selection, etc.) are dynamically adapted / adjusted based on the device configuration data 869 from the machine-learning stimulation module 860, the operating state data 877, and the control signals 881 from the control module 872.

[0127] In summary, FIGs.8A and 8B illustrate an example embodiment in which the machine- learning stimulation module 860 is configured to implement an automated learning or adaption process to learn what stimulation settings are optimal for the user (e.g., which signals and parameter settings enable the user to best perceive speech in the presence of various noisy background conditions, etc.). In certain embodiments, the machine-learning stimulation module 860 is, or includes, a classification function / model configured to generate a classification of the ambient / external sound environment, which is accordingly used to set a stimulation strategy. In other embodiments, the machine-learning stimulation module 860 is a regression / continuous function / model and the operating state data 877 comprises signal-to- noise ratio (SNR) data (e.g., channel SNR, and / or cumulative SNR at the electrode-neural interface (ENI)), that is accordingly used to set a stimulation strategy. In certain embodiments, the machine-learning stimulation module 860 includes multiple levels that perform classification and regression.

[0128] As noted, the machine-learning stimulation module 860 can be trained using a large corpus of sample data to implement the functions of the stimulation strategy adaptation module 260 of FIG.2A, including the ENI-based channel selection module 262, for example, in order to determine ENI SNR values and optimize the channel selection in a manner that results in maximizing the cumulative SNR at the electrode-neural interface. Likewise, the machine- learning stimulation module 860 can be trained using a large corpus of sample data to implement the functions of the stimulation strategy adaptation module 660 of FIG. 6A,Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 including the channel SNR estimator 658 and the ENI-based channel selection module 662, for example, in order to determine ENI SNR values and optimize the channel selection in a manner that results in maximizing the cumulative SNR at the electrode-neural interface.

[0129] According to the example embodiments described above with reference to the figures, additional gains can be provided to the cochlear implant (CI) listener by exploiting the electrode-neural interface (ENI) factors. Various techniques presented herein can minimize subject-specific SNR at the ENI, in order to provide “personalized noise reduction” for a recipient, for example.

[0130] As previously described, the technology disclosed herein can be applied in any of a variety of circumstances and with a variety of different devices, such as neurostimulators, cardiac pacemakers, cardiac defibrillators, sleep apnea management stimulators, seizure therapy stimulators, tinnitus management stimulators, and vestibular stimulation devices, as well as other medical devices that deliver stimulation to tissue. Further, technology described herein can also be applied to consumer devices. These different systems and devices can benefit from the technology described herein.

[0131] As should be appreciated, while particular uses of the technology have been illustrated and discussed above, the disclosed technology can be used with a variety of devices in accordance with many examples of the technology. The above discussion is not meant to suggest that the disclosed technology is only suitable for implementation within systems akin to that illustrated in the figures. In general, additional configurations can be used to practice the processes and systems herein and / or some aspects described can be excluded without departing from the processes and systems disclosed herein.

[0132] This disclosure described some aspects of the present technology with reference to the accompanying drawings, in which only some of the possible aspects were shown. Other aspects can, however, be embodied in many different forms and should not be construed as limited to the aspects set forth herein. Rather, these aspects were provided so that this disclosure was thorough and complete and fully conveyed the scope of the possible aspects to those skilled in the art.

[0133] As should be appreciated, the various aspects (e.g., portions, components, etc.) described with respect to the figures herein are not intended to limit the systems and processes to the particular aspects described. Accordingly, additional configurations can be used toAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 practice the methods and systems herein and / or some aspects described can be excluded without departing from the methods and systems disclosed herein.

[0134] According to certain aspects, systems and non-transitory computer readable storage media are provided. The systems are configured with hardware configured to execute operations analogous to the methods of the present disclosure. The one or more non-transitory computer readable storage media comprise instructions that, when executed by one or more processors, cause the one or more processors to execute operations analogous to the methods of the present disclosure.

[0135] Similarly, where steps of a process are disclosed, those steps are described for purposes of illustrating the present methods and systems and are not intended to limit the disclosure to a particular sequence of steps. For example, the steps can be performed in differing order, two or more steps can be performed concurrently, additional steps can be performed, and disclosed steps can be excluded without departing from the present disclosure. Further, the disclosed processes can be repeated.

[0136] Although specific aspects were described herein, the scope of the technology is not limited to those specific aspects. One skilled in the art will recognize other aspects or improvements that are within the scope of the present technology. Therefore, the specific structure, acts, or media are disclosed only as illustrative aspects. The scope of the technology is defined by the following claims and any equivalents therein.

[0137] It is also to be appreciated that the embodiments presented herein are not mutually exclusive and that the various embodiments may be combined with another in any of a number of different manners.Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 APPENDIX Mathematical Model: Equation (a): Signal model If ci(t) is the channel amplitude of the ithchannel, then ci(t) = si(t) + ni(t), where si(t) and ni(t) are the signal component and noise component of the channel amplitudes ci(t) for the ithchannel and t is the time index. Equation (b): Theoretical SNR calculation: Channel SNR of the ithchannel can be expressed as SNR(i) = ∑ (si(t)^2) / ∑ (ni(t)^2) Equation (c): Simple SOE model λ(i, j) is the dampening factor between channel i and channel j, then the contribution of ithchannel amplitude, ci(t), at the jthchannel location can be calculated as: cij(t) = λ(i, j)* ci(t) or equivalently, cij(t) = λ(i, j)* (si(t) + ni(t)) Equation (d): Estimating signal and noise amplitudes For each channel i, we will know the channel amplitudes, ci(t), and the channel based SNRs, SNRs(i), (from an SNR estimation algorithm), we can estimate an where si(t) and ni(t) from the estimated channel SNRs as follows: ^̂^^ ^^^ (i) = alpha* ci(t) and ^^^^ ^^^ (i) = (1-alpha)* ci(t) where 0<=alpha <=1, and alpha is dependent on the SNR, i.e., alpha=SNR / (1+SNR). Equation (e): Calculate the excitation pattern (ENI signal) Cumulative signal at the ENI due to all the channel amplitudes (i.e., m channels)Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 E^^t^ ൌ c^^t^ ^^λ^ ^^, ^^^ ∗ c^^t^ ^ஷ^Equation (f): ^s^^^t^ ^ ∑^ஷ^ λ^ ^^, ^^^ ∗ s^^^t^ ^ଶ^^ ^^ ^^ ^^ ^^ ^^^ ^^^ ൌ^n^ ^^ ^^ ଶ^ t ^ λ^ ^^, ^^^ ∗ n^^ t ^Cumulative SNR =∑^^ ^^ ^^ ^^ ^^ ^^^ ^^^

Claims

Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 CLAIMS What is claimed is:

1. A method, comprising: determining channelized signals from an input signal; estimating channel signal-to-noise ratios (SNRs) for a plurality of the channelized signals; determining current spread functions for at least a subset of the plurality of channelized signals; estimating, based on the current spread functions and the channel SNRs, electrode- neural interface (ENI) signal-to-noise ratios (SNRs) for at least the subset of the plurality of channelized signals; and selecting, based on the ENI SNRs, a set of the channelized signals for use in delivering stimulation signals to a recipient.

2. The method of claim 1, wherein determining current spread functions for at least a subset of the plurality of channelized signals includes: setting at least one of the current spread functions for at least one of the subset of the plurality of channelized signals based on a channel SNR of the at least one of the subset of the plurality of channelized signals.

3. The method of claim 2, wherein setting at least one of the current spread functions for at least one of the subset of the plurality of channelized signals based on the channel SNR of the at least one of the subset of the plurality of channelized signals comprises: determining that the at least one of the subset of the plurality of channelized signals has a relatively high channel SNR; and artificially increasing the current spread function of the at least one of the subset of the plurality of channelized signals.

4. The method of claim 2, wherein setting at least one of the current spread functions for at least one of the subset of the plurality of channelized signals based on the channel SNR of the at least one of the subset of the plurality of channelized signals comprises:Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 determining that the at least one of the subset of the plurality of channelized signals has a relatively low channel SNR; and artificially decreasing the current spread function of the at least one of the subset of the plurality of channelized signals.

5. The method of claim 1, 2, 3, or 4, wherein selecting, based on the ENI SNRs, a set of the channelized signals for use in delivering stimulation signals to a recipient further comprises: ensuring that a channelized signal with a highest channel SNR is included in the set of the channelized signals for use in delivering stimulation signals to the recipient.

6. The method of claim 1, 2, 3, or 4, wherein selecting, based on the ENI SNRs, a set of the channelized signals for use in delivering stimulation signals to a recipient further comprises: enforcing a predetermined channel independence parameter to impose a minimum separation or distance between selected channels.

7. The method of claim 6, wherein the predetermined channel independence parameter is an electrode-based spacing parameter.

8. The method of claim 6, wherein the predetermined channel independence parameter is a region-based spacing parameter.

9. The method of claim 1, 2, 3, or 4, wherein the input signals are environmental signals.

10. The method of claim 9, wherein the input signals are sound signals.

11. A processing unit for processing a spatial signal, comprising: a filter bank configured to process the spatial signal to generate channel signals in each of a plurality of spaced frequency channels, wherein each channel is associated with at least one electrode of a plurality of electrodes configured to be implanted in a recipient;Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 a channel signal-to-noise ratio (SNR) estimator configured to estimate channel SNRs associated with the channel signals; and an electrode-neural interface (ENI)-based channel selection module configured to: calculate, a vector of SNRs at an interface between one or more of the plurality of electrodes and neurons based on the channel SNRs, and select, based at least in part on the vector SNRs at the interface between the one or more of the plurality of electrodes and the neurons, a subset of the channel signals for use in stimulating the recipient based on the spatial signal.

12. The processing unit of claim 11, further comprising: a current spread function module configured to model spread of excitation (SOE) patterns associated with the channel signals, wherein the ENI-based channel selection module is configured to select the subset of the channel signals further based on the SOE patterns modeled by the current spread function module.

13. The processing unit of claim 12, wherein the current spread function module is configured to adjust one or more of the SOE patterns associated with one or more of the channel signals based on the channel SNRs.

14. The processing unit of claim 13, wherein to adjust one or more of the SOE patterns associated with one or more of the channel signals based on the channel SNRs, the current spread function module is configured to: adaptively increase an SOE pattern associated with a channel signal having a relatively high channel SNR associated therewith.

15. The processing unit of claim 14, wherein to adjust one or more of the SOE patterns associated with one or more of the channel signals based on the channel SNRs, the current spread function module is configured to: adaptively decrease an SOE pattern associated with a channel signal having a relatively low channel SNR associated therewith.

16. The processing unit of claim 11, 12, 13, 14, or 15, wherein the ENI-based channel selection module is configured to select the subset of the channel signals that maximizes theAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 vector of SNRs at the interface between the one or more of the plurality of electrodes and the neurons.

17. The processing unit of claim 11, 12, 13, 14, or 15, wherein the ENI-based channel selection module is configured to select the subset of the channel signals that maximizes channel independence between selected channel signals according to a predetermined channel spacing parameter.

18. The processing unit of claim 11, 12, 13, 14, or 15, wherein to select the subset of the channel signals for use in stimulating the recipient based on the spatial signal, the ENI-based channel selection module is configured to: identify a first channel signal having a highest channel SNR associated therewith; and select at least the first channel signal having the highest channel SNR associated therewith for inclusion in the subset of the channel signals for use in stimulating the recipient.

19. The processing unit of claim 18, wherein the ENI-based channel selection module is further configured to: identify a second channel signal having a second highest channel SNR associated therewith; and select at least the second channel signal having the second highest channel SNR associated therewith for inclusion with the first channel signal in the subset of the channel signals for use in stimulating the recipient.

20. The processing unit of claim 11, 12, 13, 14, or 15, wherein the spatial signal is a sound signal.

21. A method comprising: obtaining a plurality of channel signals corresponding to an environmental signal; estimating local channel signal-to-noise ratios (SNRs) for each of the plurality of channel signals; calculating cumulative SNRs at an electrode-neural interface (ENI) for different subsets of channel signals among the plurality of channel signals based on the local channel SNRs; andAtty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 selecting one of the different subsets of the plurality of channel signals for use in stimulation at least based on the cumulative SNRs at the ENI.

22. The method of claim 21, further comprising: adjusting channel selection for stimulation based on the cumulative SNRs at the ENI by selecting a subset of channel signals that increases the cumulative SNRs at the ENI.

23. The method of claim 21, further comprising: adjusting channel selection for stimulation based on the cumulative SNRs at the ENI by selecting a subset of channel signals that increases separation between actively stimulating electrodes corresponding to respective channel signals.

24. The method of claim 23, wherein increasing the separation between the actively stimulating electrodes reduces overlap of current spread patterns at the ENI, thereby increasing channel independence of the actively stimulating electrodes.

25. The method of claim 23, wherein increasing the separation between the actively stimulating electrodes reduces interference from a neighboring channel having a low SNR, thereby increasing the cumulative SNR at the electrode-neural interface.

26. The method of claim 21, 22, 23, 24, or 25, further comprising: adjusting channel selection for stimulation based on the cumulative SNRs at the ENI using a greedy search maxima selection algorithm to select n maxima out of m channels.

27. The method of claim 26, wherein the greedy search maxima selection algorithm is configured to select the n maxima out of the m channels so as to maximize the cumulative SNRs at the ENI.

28. The method of claim 26, wherein the greedy search maxima selection algorithm is configured to select the n maxima out of the m channels so as to maximize channel independence between actively stimulating electrodes corresponding to the selected channel signals.Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 29. The method of claim 26, wherein the greedy search maxima selection algorithm is configured to select the n maxima out of the m channels so as to balance the cumulative SNRs at the ENI with channel independence between actively stimulating electrodes corresponding to the selected channel signals based on a metric.

30. The method of claim 21, 22, 23, 24, or 25, further comprising: determining a spread of excitation (SOE) pattern per channel for each of the plurality of channel signals using a current spread function per channel; identifying one or more channel signals having a relatively high local channel SNR from among the plurality of channel signals; and adaptively increasing the SOE pattern for the one or more channel signals having the relatively high local channel SNR.

31. The method of claim 30, wherein adaptively increasing the SOE pattern for the one or more channel signals having the high local channel SNR assigns a larger weight to the one or more channel signals having the high local channel SNR in the current spread function, thereby increasing a likelihood that the one or more channel signals having the relatively high local channel SNR will be selected for use in stimulation.

32. The method of claim 21, 22, 23, 24, or 25, further comprising: adaptively adjusting channel selection for stimulation based on amplitudes of the plurality of channel signals, a modeled spread of excitation (SOE) pattern determined using a current spread function, and the local channel SNRs for the plurality of channel signals.

33. The method of claim 32, wherein adaptively adjusting channel selection for stimulation based on the amplitudes, the modeled SOE pattern, and the local channel SNRs comprises: deselecting one or more channel signals having a relatively low local channel SNR; and selecting one or more channel signals having a relatively high local channel SNR.

34. The method of claim 21, 22, 23, 24, or 25, wherein calculating the cumulative SNRs at the ENI comprises:Atty. Docket No.3065.0728i Client Ref. No. CID03577WOPC1 calculating a vector of local channel SNRs observed at corresponding neural elements of the ENI for a given current spread function (SOE pattern).

35. The method of claim 34, further comprising: dynamically adjusting channel selection for stimulation based on the vector of local channel SNRs observed at the corresponding neural elements of the ENI such that the one of the different subsets of the plurality channel signals that is selected for stimulation maximizes the cumulative SNR at the ENI.

36. The method of claim 35, wherein dynamically adjusting channel selection for stimulation based on the vector of local channel SNRs observed at the corresponding neural elements of the ENI comprises: selecting one of the different subsets of the plurality of channel signals for use in stimulation based on a contribution of signal to a measured spread of excitation (SOE) pattern.

Images