System, method and device for cancelling a perception of a user's own voice

Abstract

The invention relates to adaptive signal processing for human voice control and acoustic feedback management. More particularly, it concerns a system and method for detecting and cancelling a person's own voice, as may be perceived through airborne, bone, and soft-tissue conduction. The invention also covers related calibration, adaptive filtering, and neural-network implementations for such systems. The present invention represents a system and method for cancelling a perception of a user's own voice as perceived through airborne, bone, and soft-tissue conduction paths. The suggested solution may operate with minimal latency, adapt to individual anatomical and environmental conditions, and maintain transparency to external sounds. Furthermore, the solution incorporates calibration routines to accurately model the transmission characteristics of these paths. It also allows to employ adaptive or machine-learning techniques to refine cancellation performance in real time.

Description

SYSTEM, METHOD AND DEVICE FOR CANCELLING A PERCEPTION OF A USER’S OWN VOICECROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of priority of U.S. Provisional Patent Application No. 63 / 734,906, filed December 17, 2024, and entitled: “System and Method for person's own-hearing cancellation and applications thereof’ which is hereby incorporated by reference in its entirety.FIELD OF THE INVENTION

[0002] The invention relates to adaptive signal processing for human voice control and acoustic feedback management. More particularly, it concerns a system and method for detecting and cancelling a person’s own voice, as may be perceived through airborne, bone, and soft-tissue conduction. The invention also covers related calibration, adaptive filtering, and neural-network implementations for such systems.BACKGROUND OF THE INVENTION

[0003] When a person speaks, they hear their own voice not only through the air but also via internal transmission through the skull and soft tissue. This self-hearing, or perception of a person’s own voice (POV), is a combined result of air conduction, bone conduction, and tissue conduction.

[0004] In normal conditions this blend feels natural; however, in amplified or occluded situations (such as with hearing aids or enclosed headsets), it leads to discomfort and distortion known as the occlusion effect.

[0005] Hearing-aid users often report their own voice sounding louder, “boomy,” or “in the head.” Speech-therapy patients using delayed auditory feedback (DAF) or frequency-altered feedback (FAF) devices may experience interference between the natural self-voice and the modified feedback. Public speakers or performers using monitors often perceive confusing colouration of their own voice through feedback paths.

[0006] In all these cases, the self-voice component is highly correlated with the speaker’s actual vocal signal and propagates through multiple coupled channels with different delays and transfer functions.

[0007] Traditional audio cancellation techniques optimized for environmental noise cannot cancel this internal component, because a portion of the self-voice originates inside the speaker’s head, not in the surrounding air.

[0008] The process of hearing our own voice is a fundamental one. As long as we can hear, we hear ourselves. One cannot stop hearing himself even if he attempts to block the input to his ears, for example by inserting earplugs in the ear canals. With earplugs it is possible to hear yourself even stronger.

[0009] Delayed Auditory Feedback (DAF) is a process in which a replica of the POV is delayed in time and then played to the person’s ear. It is known in the art that DAF can improve speech fluency of stuttering people. Many people who stutter experience improved speech fluency when they hear their POV slightly delayed. Typically, delays are of the order of 0.1 -0.2 seconds.

[0010] Frequency Altered Feedback (FAF), is a process of modifying the frequencies of the speech (without changing its speed). The speech fluency of some people who stutter improves when they hear POV with modified frequency components.

[0011] DAF and FAF are referred to generally as Modified Audio Feedback or MAF.

[0012] Some users who stutter or clutter experience further speech improvement when applying DAF and FAF together.

[0013] Additionally, and optionally, the actual cancellation of a person’s own hearing of their own voice may have beneficial effects by itself, with no additional injection of sound, for stuttering, cluttering and stage fright.

[0014] Current applications of DAF or FAF which superimpose the natural own hearing can cause discomfort. Some users require training to accustom themselves to using the MAF devices; some, although their speech fluency is improved, eventually stop using the devices because of this uneasiness. There is also a known phenomenon whereby the favorable effects of DAF or FAF diminish with extended use. All these are most probably related to the presence of the natural POV in addition to the modified signal. The elimination of the POV is expected to have a positive effect on the performance of DAF and FAF and alleviate the discomfort and allow prolonged use of the devices.

[0015] As known, about 1% of the population are people who stutter; and also about 1% are people who clutter. About 30% of each group both stutter and clutter. Many others have stage fright, Alzheimer, aphasia and dysarthria and fear of speaking in public. There areindications that control and manipulation of POV can assist in overcoming these conditions. Other potential applications include the cancellation of the POV on its own, improved crosstalk performance in audiometric tests, a certain relief of tinnitus, and various other audio- psychological disorders as further described in the following.

[0016] Active sound cancellation techniques have traditionally focused on reducing unwanted environmental noise or filtering background sounds from speech transmission. These systems typically operate by generating an anti-phase signal to attenuate external acoustic disturbances, thereby improving the clarity of communication or reducing ambient noise for the user. Common applications include headsets, hearing aids, and telecommunication devices, where the primary goal is to suppress external noise sources rather than signals originating from the user.

[0017] Other known approaches have been directed toward limiting sound leakage from devices, such as preventing speech from being audible to nearby individuals during phone calls or in shared environments. These solutions aim to enhance privacy and minimize disturbance to surrounding personnel by emitting counter-phase signals to cancel sound projected outward from the user.

[0018] While such techniques are effective for external noise or sound leakage, they do not address the unique challenge of cancelling a person’s own voice as perceived internally. When speaking, a user hears his own voice through multiple transmission paths, including airborne sound, bone conduction, and soft-tissue conduction. These internal pathways introduce complex delays, phase shifts, and spectral coloration that differ significantly from external noise characteristics. Conventional active noise control systems are not designed to compensate for these internal propagation effects, nor do they synchronize cancellation with the transient nature of speech signals. As a result, existing solutions cannot effectively eliminate the discomfort or distortion associated with self-hearing, particularly in occluded or amplified conditions such as hearing aids, enclosed headsets, or therapeutic devices.SUMMARY OF THE INVENTION

[0019] Accordingly, there is a need for a system and method for cancelling a perception of a user’s own voice as perceived through airborne, bone, and soft-tissue conduction paths. Such a solution should operate with minimal latency, adapt to individual anatomical and environmental conditions, and maintain transparency to external sounds. Furthermore, it should incorporate calibration routines to accurately model the transmission characteristicsof these paths and optionally employ adaptive or machine-learning techniques to refine cancellation performance in real time.

[0020] To address the aforementioned needs, the following is suggested.

[0021] The present invention provides a multi-channel adaptive system that may detect a user’s own voice and generate a corresponding cancellation signal. Sensors may be configured to detect the self-voice along the airborne, and optionally in the bone-conduction, and soft-tissue paths. A processing module may be configured to compute an inverse signal with each of its spectral components matched in amplitude, phase, and delay. This inverse signal may be emitted by transducers positioned near the ear to essentially cancel the perceived self-voice.

[0022] In one embodiment the cancellation may be based on estimating each path’s transfer function H(f) (also may be referred herein as TF) and generating a signal with frequency response -H(f). These transfer functions may include any environmental effects, such as echoes or reverberations. In other implementations the cancellation signal may be generated with machine learning tools.

[0023] Residual-signal feedback from a monitoring sensor near the ear can be used to refine the adaptive filters for improved cancellation. To improve spectral precision, the system may dynamically adjust its sampling window and / or sampling frequency so that FFT frequency bins align with the strongest frequency components of the POV.

[0024] In a general aspect the invention may be directed to a system for cancelling a perception of a user’s own voice (POV), comprising: a voice input sensor configured to detect a source voice signal generated by the user; a cancellation transducer positioned proximate to the user's ear; and a processing module configured to: receive the source voice signal; generate a cancellation signal by applying a transfer function (TF) to the source voice signal, wherein the transfer function models a combined acoustic transmission path of air, bone, and soft-tissue from the user's mouth to the user's ear; and drive the transducer with the cancellation signal to destructively interfere with the user’s own voice.

[0025] In some embodiments, the processing module may be configured to estimate a boneconduction component and a soft-tissue component of the user's voice based solely on the airborne source voice signal detected by the voice input sensor, and combine these estimated components into the cancellation signal.

[0026] In some embodiments, the processing module may be configured to calculate the cancellation signal one sample at a time using a sliding input window, such that a processing latency of the system may be lower than a time-of-flight of the user's voice from the mouth to the ear.

[0027] In some embodiments, the processing module may include a Neural Network trained to generate the cancellation signal from an input sampling window that is shorter than a period of a lowest frequency component of the user's voice, thereby compensating for a short acoustic delay of the transmission path.

[0028] In some embodiments, the processing module may be configured to forecast the cancellation signal for a future time instance to compensate for hardware and software latencies.

[0029] In some embodiments, the cancellation transducer may include at least one of an air-conduction speaker, a bone-conduction exciter, or a direct electrical interface to a hearing aid.

[0030] In some embodiments, the system may further include a monitoring microphone at the user's ear, wherein the processing module calibrates the transfer function (TF) using the user's natural spoken voice by calculating a ratio of a signal at the monitoring microphone (VE) to a signal at the voice input sensor (VM), normalized by a transfer function of the monitoring microphone and cancellation transducer (TFEE).

[0031] In some embodiments, the transfer function may be derived from an offline calibration process comprising: measuring a mouth-to-ear response (TFME), measuring a mouth-to-mouth response (TFMM), and measuring a cancellation-hardware-to-ear response (TFCE), and calculating the transfer function as TF = - TFME / (TFMM • TFCE).

[0032] In another general aspect, the present invention may be directed to a method for realtime cancellation of a user's self-hearing, comprising: detecting a voice signal generated by the user; transmitting a cancellation signal to the user's ear based on a transfer function; monitoring a residual signal at the user's ear; and updating the transfer function in real-time by subtracting a correction factor, wherein said correction factor is derived from the residual signal normalized by a transfer function of an ear-speaker and ear-microphone path.

[0033] In some embodiments, the method may further include dynamically adjusting a sampling window size or sampling frequency to align frequency bins with dominant spectral components of the user's voice.

[0034] In some embodiments, the method may further include generating a Modified Audio Feedback (MAF) signal, comprising at least one of a time delay (DAF) or frequency shift (FAF), and superimposing the MAF signal with the cancellation signal to treat speech fluency disorders.

[0035] In some embodiments, the method may be applied to a hearing aid, wherein the cancellation signal may be configured to reduce an occlusion effect caused by the hearing aid blocking an ear canal of the user.

[0036] In some embodiments, the method may further include detecting eardrum vibrations associated with tinnitus and generating the cancellation signal to suppress a perceived tinnitus sound.

[0037] In another general aspect, the present invention may be directed to a non-transitory computer-readable medium comprising instructions to: detect an audio signal at an ear-canal microphone; process the audio signal using a Neural Network trained to identify a specific voice signature of a user; and generate a cancellation signal only when the specific voice signature is identified, thereby preserving environmental sounds.

[0038] In some embodiments, the instructions may be configured to generate the cancellation signal based solely on input from the ear-canal microphone without input from a mouth-proximate microphone.

[0039] In yet another general aspect, the present invention may be directed to a device for cancelling a user’s perception of a user’s own voice (POV), comprising: at least one voice input sensor configured to detect a source signal associated with the user’s voice; at least one cancellation transducer configured to emit an acoustic cancelling signal; and at least one non-transitory memory device storing modules of instruction code; and and at least one processor operatively coupled to the at least one memory device, the at least one voice input sensor, and the at least one cancellation transducer, and being configured to execute the modules of instruction code, whereupon execution of said modules of instruction code, the at least one processor is configured to: receive the source signal from the at least one voice input sensor; generate the acoustic cancelling signal by applying a pre-configured transfer function (TF) representing a function of: (a) the source signal detected by the at least one voice input sensor, and (b) a target signal obtained after passing the source signal through an acoustic transmission path including one or more of: (i) an air channel; (ii) a bone channel through a bone tissue of the user; and (iii) a soft-tissue channel through a soft tissue of theuser, said acoustic cancelling signal being configured to destructively interfere with the target signal; and drive the at least one cancellation transducer with the acoustic cancelling signal.

[0040] In some embodiments, the at least one voice input sensor may include at least one of (a) a microphone; (b) a bone-conduction sensor; and (c) a soft-tissue vibration sensor; and the at least one cancellation transducer may include at least one of (i) an air-conduction speaker; (ii) a bone-conduction exciter; and (iii) a soft-tissue transducer.

[0041] In some embodiments, the at least one processor may be further configured to modify the source signal by applying a time delay or a frequency shift thereto, and drive the at least one cancellation transducer with the acoustic cancelling signal while superimposing the modified source signal onto the acoustic cancelling signal.

[0042] In some embodiments, the pre-configured transfer function (TF) may be derived from an offline calibration process comprising: computing a mouth-to-ear transfer function as a function of (a) a first reference signal as transmitted through a configuration transducer positioned proximate to the at least one voice input sensor; and (b) a first detected signal as obtained by detecting the first reference signal by a configuration sensor positioned proximate to the at least one cancellation transducer; computing a cancelling-channel transfer function as a function of: (a) a second reference signal as transmitted through the at least one cancellation transducer positioned proximate to the configuration sensor; and (b) a second detected signal as obtained by detecting the second reference signal by the configuration sensor; and computing a mouth-channel transfer function as a function of: (a) a third reference signal as transmitted through the configuration transducer positioned proximate to the at least one voice input sensor; and (b) a third detected signal as obtained by detecting the third reference signal by the at least one voice input sensor; and calculating the TF in the frequency domain as a function of the mouth-to-ear transfer function; the cancelling-channel transfer function and the mouth-channel transfer function.

[0043] In some embodiments, the TF may be calculated in the frequency domain as an inverted Hadamard element-wise division of the mouth-to-ear transfer function by the product of the mouth-channel transfer function and the cancelling-channel transfer function.

[0044] In some embodiments, the at least one processor may be configured to generate the acoustic cancelling signal by: transforming the source signal into a spectral representation in the frequency domain; applying the TF to the spectral representation to compute acancelling signal spectrum, wherein each spectral component of the cancelling signal spectrum is amplitude- and phase-adjusted so as to compensate for delay and reflections of the acoustic transmission path; converting the cancelling signal spectrum into a time-domain by performing an inverse Fourier Transform, thereby obtaining the acoustic cancelling signal; and synchronizing the acoustic cancelling signal with an estimated arrival of the user’s own voice at the user’s ear to achieve destructive interference.

[0045] In some embodiments, the device may further include a residual signal correction sensor operatively coupled to said at least one processor and positioned in proximity to the at least one cancellation transducer sensor; and wherein said at least one processor may be further configured to: receive an input signal from the residual signal correction sensor upon driving the at least one cancellation transducer with the acoustic cancelling signal, said input signal representing an un-cancelled portion of the user’s own voice; determine the residual signal in the frequency domain as a function of the input signal combined with the TF and an ear-channel transfer function pre-computed as a function of: (a) a fourth reference signal as transmitted through the at least one cancellation transducer; and (b) a fourth detected signal as obtained by detecting the fourth reference signal by the residual signal correction sensor; update the TF based on the determined residual signal, thereby generating a corrected transfer function; and further generating the acoustic cancelling signal by applying the corrected transfer function, so as to reduce the residual signal.

[0046] In yet another general aspect, the present invention may be directed to a method for cancelling a user’s perception of a user’s own voice (POV), by at least one processor, comprising: receiving a source signal associated with the user’s voice from at least one voice input sensor; generating an acoustic cancelling signal by applying a pre-configured transfer function (TF) representing a function of: (a) the source signal detected by the at least one voice input sensor, and (b) a target signal obtained after passing the source signal through an acoustic transmission path including one or more of: (i) an air channel; (ii) a bone channel through a bone tissue of the user; and (iii) a soft-tissue channel through a soft tissue of the user, said acoustic cancelling signal being configured to destructively interfere with the target signal; and driving at least one cancellation transducer with the acoustic cancelling signal.BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The subject matter regarded as the invention is particularly pointed out and distinctly claimed in the concluding portion of the specification. The invention, however, both as to organization and method of operation, together with objects, features, and advantages thereof, may best be understood by reference to the following detailed description when read with the accompanying drawings in which:

[0048] Fig. 1 is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating the three acoustic transmission channels through which a person’s own voice (POV) propagates - airborne, bone-conduction, and soft-tissue conduction - and showing sensors and transducers positioned to detect and cancel the POV signal, according to some embodiments of the present invention.

[0049] Fig. 2 is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an embodiment incorporating a Modified Audio Feedback (MAF) module configured to apply a time delay or frequency shift to the detected voice signal and superimpose the modified signal with the cancelling signal for therapeutic applications, according to some embodiments of the present invention.

[0050] Fig. 3 is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating a configuration in which the system implements two cancelling channels — airborne and bone-conduction — while deriving soft-tissue cancellation from the bone-conduction path, according to some embodiments of the present invention.

[0051] Fig. 4 is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating a simplified configuration implementing only the airborne cancelling channel, wherein the cancelling signal is emitted through an air-conduction transducer positioned proximate to the ear, according to some embodiments of the present invention.

[0052] Fig. 5A is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an offline calibration arrangement for determining the mouth-to-ear transfer function by transmitting a reference signal through a mouth speaker and detecting the response at an ear microphone, according to some embodiments of the present invention.

[0053] Fig. 5B is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an offline calibration arrangement for determining the cancelling-channel transfer function by transmitting a reference signal through a cancellation speaker and detecting the response at the ear microphone, according to some embodiments of the present invention.

[0054] Fig. 5C is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an offline calibration arrangement for determining the mouth-channel transfer function by transmitting a reference signal through the mouth speaker and detecting the response at a mouth microphone, according to some embodiments of the present invention.

[0055] Fig. 5D is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an arrangement for demonstrating cancellation of airborne voice transmission by simultaneously transmitting an audio signal through the mouth speaker and a cancelling signal through the cancellation speaker, according to some embodiments of the present invention.

[0056] Fig. 6A is a flow chart illustrating a calibration process for determining the transfer function of the mouth speaker to the mouth microphone (MS-MM channel), including Fourier transformation of detected signals, according to some embodiments of the present invention.

[0057] Fig. 6B is a flow chart illustrating a calibration process for determining the transfer function of the ear speaker to the ear microphone (EE-EM channel), including storage of the calculated transfer function for later use, according to some embodiments of the present invention.

[0058] Fig. 6C is a flow chart illustrating a calibration process for determining the transfer function of the mouth speaker to the ear microphone (MS-EM channel) and combining previously calculated transfer functions to compute the overall transfer function of the system, according to some embodiments of the present invention.

[0059] Fig. 6D is a flow chart illustrating a process for measuring and compensating for delay and environmental reflections affecting the cancelling signal, including synchronization of the cancelling signal with the arrival of the POV at the ear, according to some embodiments of the present invention.

[0060] Fig. 7 is a flow chart illustrating a real-time cancellation process using a precalibrated transfer function and monitoring of any residual signal at the ear, wherein the residual signal is used to modify and update the transfer function for improved cancellation, according to some embodiments of the present invention.

[0061] Fig. 8 A is a schematic illustration of a head-mounted sound module including an airchannel input microphone, an air-cancelling speaker, and a bone-conduction transducer, configured for placement over the ear, which may be integrated into a system or device for cancelling a perception of a user’ s own voice, according to some embodiments of the present invention.

[0062] Fig. 8B is a schematic illustration of a head-mounted configuration with two sound modules of a system or device for cancelling a perception of a user’s own voice, secured by a head strap, each module including components for air and bone-conduction cancellation, according to some embodiments of the present invention.

[0063] Fig. 9 is a flow chart illustrating an implementation of the transfer function as a finite impulse response (FIR) filter in the time domain, including convolution-based generation of the cancelling signal, according to some embodiments of the present invention.

[0064] Fig. 10 is a flow diagram depicting a method for cancelling a perception of a user’s own voice, according to some embodiments of the present invention.

[0065] It will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity. Further, where considered appropriate, reference numerals may be repeated among the figures to indicate corresponding or analogous elements.DETAILED DESCRIPTION OF THE PRESENT INVENTION

[0066] One skilled in the art will realize the invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects illustrative rather than limiting of the invention described herein. Scope of the invention is thus indicated by the appended claims, rather than by the foregoing description, and all changes that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[0067] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be understood by those skilled in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the present invention. Some features or elements described with respect to one embodiment may be combined with features or elements described with respect to other embodiments. For the sake of clarity, discussion of same or similar features or elements may not be repeated.

[0068] Although embodiments of the invention are not limited in this regard, discussions utilizing terms such as, for example, “processing,” “computing,” “calculating,” “determining,” “establishing”, “analyzing”, “checking”, “choosing”, “selecting”, “omitting”, “training”, “applying”, “forming” or the like, may refer to operation(s) and / or process(es) of a computer, a computing platform, a computing system, or other electronic computing device, that manipulates and / or transforms data represented as physical (e.g., electronic) quantities within the computer’s registers and / or memories into other data similarly represented as physical quantities within the computer’s registers and / or memories or other information non-transitory storage medium that may store instructions to perform operations and / or processes.

[0069] Although embodiments of the invention are not limited in this regard, the terms “plurality” and “a plurality” as used herein may include, for example, “multiple” or “two or more”. The terms “plurality” or “a plurality” may be used throughout the specification to describe two or more components, devices, elements, units, parameters, or the like. The term “set” when used herein may include one or more items.

[0070] Unless explicitly stated, the method embodiments described herein are not constrained to a particular order or sequence. Additionally, some of the described method embodiments or elements thereof can occur or be performed simultaneously, at the same point in time, concurrently, or iteratively and repeatedly.

[0071] In embodiments of the present invention, some steps of the claimed method may be performed using machine-learning (ML)-based models or may include actions performed on ML-based models, e.g., transferring ML-based models over a computer network. ML- based models may be configured or “trained” for a specific task, e.g., classification or regression.

[0072] In some embodiments, ML-based models may be artificial neural networks (ANN).

[0073] A neural network (NN) or an artificial neural network (ANN), e.g., a neural network implementing a machine learning (ML) or artificial intelligence (Al) function, may refer to an information processing paradigm that may include nodes, referred to as neurons, organized into layers, with links between the neurons. The links may transfer signals between neurons and may be associated with weights. A NN may be configured or trained for a specific task, e.g., pattern recognition or classification. Training a NN for the specific task may involve adjusting these weights based on examples. Each neuron of an intermediate or last layer may receive an input signal, e.g., a weighted sum of output signals from other neurons, and may process the input signal using a linear or nonlinear function (e.g., an activation function). The results of the input and intermediate layers may be transferred to other neurons and the results of the output layer may be provided as the output of the NN. Typically, the neurons and links within a NN are represented by mathematical constructs, such as activation functions and matrices of data elements and weights. A processor, e.g., CPUs or graphics processing units (GPUs), or a dedicated hardware device may perform the relevant calculations.

[0074] It should be obvious for the one ordinarily skilled in the art that various ML-based models can be implemented without departing from the essence of the present invention. It should also be understood, that in some embodiments ML-based model may be a single ML- based model or a set (ensemble) of ML-based models realizing as a whole the same function as a single one. Hence, in view of the scope of the present invention, the abovementioned variants should be considered equivalent.

[0075] According to the concept of the present invention, the technical approach fundamentally differs from conventional Active Noise Control (ANC) systems. While ANC solutions are designed to cancel external environmental noise by generating an anti-phase signal based on omnidirectional microphones, they do not address the unique characteristics of a person’s own voice (POV), which: (a) originates internally and propagates through multiple coupled paths - airborne, bone-conduction, and soft-tissue conduction. Exhibits complex delays, phase shifts, and spectral coloration that vary dynamically and differ from external noise. Requires synchronous cancellation of transient speech signals with extremely low latency.

[0076] The present invention solves these problems through the following. The system computes a transfer function (TF) representing the combined mouth-to-ear acoustic path, including air, bone, and tissue channels. This enables precise amplitude and phase inversion for each spectral component, unlike ANC which assumes a single external air path. The processing module generates the cancellation signal sample-by-sample using a sliding window, ensuring latency lower than the mouth-to-ear time-of-flight. Conventional ANC tolerates higher latency because it targets stationary noise. The invention optionally employs a Neural Network (NN) trained to forecast the user’s voice waveform and compensate for hardware / software delays. This predictive capability is absent in ANC systems. A monitoring microphone near the ear measures residual signals and updates the TF in real time, maintaining cancellation accuracy under changing conditions. ANC systems do not incorporate such adaptive feedback for internal voice cancellation. Unlike ANC, which often suppresses all incoming sound, the claimed system selectively cancels only the user’s own voice while preserving environmental audio, ensuring natural listening.

[0077] Fig. 1 is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating the three acoustic transmission channels through which a person’s own voice (POV) propagates - airborne, bone-conduction, and soft-tissue conduction - and showing sensors and transducers positioned to detect and cancel the POV signal, according to some embodiments of the present invention.

[0078] According to some embodiments of the invention, the suggested system and device may be implemented as a combination of hardware and software modules. Components of the system and device may be adapted to execute one or more modules of instruction code to request, receive, analyze, calculate and produce various data.

[0079] As further described in detail herein, the suggested system and device may be adapted to execute one or more modules of instruction code in order to perform steps of the claimed method.

[0080] As shown in Figs. 1-4, arrows may represent flow of one or more data elements to and from the system and device and / or among modules or elements of the system and device. Some arrows may have been omitted in Figs. 1-4 for the purpose of clarity.

[0081] Fig. 1, illustrates schematically the three channels through which a person’s selfhearing of the person’s own voice (POV), 1, are transmitted. Speech originates in the vocal tract, and mouth and generated to the ears in three channels: (a) the air channel, 4, where thevoice reaches the eardrum, 5, through the ear canal (not shown in Fig. 1); (b) the bone channel, 6, where the voice is transmitted through bones (also referred herein as “bode tissue”) in the skull directly to the inner ear, 7; and (c) the soft-tissue channel, 8, where the voice is transmitted through the soft-tissue of the head towards the different components of the ear, including the ear canal (not shown in Fig. 1), the ear drum, 5, the middle ear 9, and the inner ear, 7. Each of these acoustic channels is characterized by different parameters of spectral attenuation and delay.

[0082] In some embodiments, suitable sensors, 10, 11, 12, detect the voice in the aforementioned channels. Their outputs applied to adaptive filters 13, 14, 15, judiciously designed for each of the voice channels for the purpose of cancelling the contribution of that channel as described below. Each of the signals may be then inverted by an inverter, 16. Each such processed signal may be then fed to the cancelling air, 34, bone, 36, and the soft- tissue 38 channels. For this purpose, suitable transducers may be used, which are broadly referred to here as an earphone transmitter (also referred herein as “air-conduction speaker”), 24, bone transmitter (also referred herein as “bone-conduction exciter”), 26, and soft-tissue transmitter (also referred herein as “soft-tissue transducer”), 28. The variety of alternative devices that can be used to implement these transducers according to the current invention are described in the following. As can be deducted from the schematic description in Fig. 1, these three cancelling signals may be then fed back into the air channel to reach the eardrum, 5, the middle ear, 9, and the inner ear 7, being suitably weighted and delayed so as to cancel each of the POV signal channel components reaching the same ear. Each of the filters, 13, 14, 15 may be designed to ensure that each spectral component from the detected voice signal will be suitably delayed to account for the total delay in the channel (including the delay in the channel up to the sensor, 10, 11, or 12, the filters 13, 14, 15 the inverters, 16, the transducers 24, 26, 28 and the delay in transmission to the ear components) as well as the total attenuation or gain of each spectral component (including the attenuation in the channel up to the sensor, 10, 11, or 12, the filters 13, 14, 15, inverters, 16, the transducers 24, 25, 26 and the delay in transmission to the ear components), all in order to ensure that the cancelling signal substantially cancels the voice signal in each channel. Fig. 1 also shows the presence of sound originating from the surrounding (environment sounds, 40), including speech of others, which the current invention proposes to maintain with minimal interruption.

[0083] In some embodiments, an earphone or receiving sensor unit may include one or more sensors, each providing an electrical output signal (such as voltage or current), or changing a characteristic (such as resistance or impedance) in response to acoustic vibrations, or sound. The sensors may be identical, similar or different from each other, and may measure or detect sound in the same or different characteristics, such as direction, amplitude, acceptance angle or frequency content. The sensors may be optimized for the measurement or detection of airborne or tissue-borne or bone-borne sound. Two or more sensors may be connected in series or in parallel for the purpose of improving the sensing performance of each channel. In the case of a changing characteristic sensor or in the case of an active sensor, the unit may include an excitation or measuring circuits (such as a bridge) to generate the sensor electrical signal. The sensor output signal may be conditioned by a signal conditioning circuit. The signal conditioner may involve time, frequency, or magnitude related manipulation. The signal conditioner may be linear or non-linear, and may include an operation or an instrument amplifier, a multiplexer, a frequency converter, a frequency - to-voltage converter, a voltage-to-frequency converter, a current-to-voltage converter, a current loop converter, a charge converter, an attenuator, a sample-and-hold circuit, a peakdetector, a voltage or current limiter, a delay line or circuit, a level translator, a galvanic isolator, an impedance transformer, a linearization circuit, a calibrator, a passive or active (or adaptive) filter, an integrator, a deviator, an equalizer, a spectrum analyzer, a compressor or a de-compressor, a coder (or decoder), a modulator (or demodulator), a pattern recognizer, a smoother, a noise remover, an average or RMS circuit, or any combination thereof. In the case of analog sensor, an analog to digital (A / D) converter may be used to convert the conditioned sensor output signal to a digital sensor data. The unit may include a processor or computer for controlling and managing the unit operation, processing the digital sensor data and handling the unit communication. The unit may include a modem or transceiver coupled to a network port (such as a connector or antenna), for interfacing and communicating over a network.

[0084] In some embodiments, a speaker (e.g., an air-conduction speaker), transmitting transducer or actuator unit may include one or more actuators, each generating acoustic waves or sound in response to an electronic command, which can be an electronic signal (such as voltage or current), or by changing a characteristic (such as resistance or impedance) of a device. The actuators may be identical, similar or different from each other, and maygenerate sound of the same or different characteristics, such as direction, amplitude, acceptance angle or frequency content. Two or more actuators may be connected in series or in parallel to improve the performance of each channel. The actuator command signal may be conditioned by a signal conditioning circuit. The signal conditioner may involve time, frequency, or magnitude related manipulations. The signal conditioner may be linear or non-linear, and may include an amplifier, a voltage or current limiter, an attenuator, a delay line or circuit, a level translator, a galvanic isolator, an impedance transformer, a linearization circuit, a calibrator, a passive or active (or adaptive) filter, an integrator, a deviator, an equalizer, a spectrum analyzer, a compressor or a de-compressor, a coder (or decoder), a modulator (or demodulator), a pattern recognizer, a smoother, a noise remover, an average or RMS circuit, or any combination thereof. In the case of analog actuator, a digital to analog (D / A) converter may be used to convert the digital command data to analog signals for controlling the actuators. The unit may include a processor for controlling and managing the unit operation, processing the actuators commands and handling the unit communication. The unit may include a modem or transceiver coupled to a communication port (such as a connector or antenna), for interfacing and communicating over a network.

[0085] In some embodiments, a sensor / actuator unit may be a device integrating a part or whole of a sensor unit with part or whole of an actuator unit. For a non-limiting example, such hardware integration may relate to housing in the same enclosure, sharing the same connector (power, communication or any other connector), sharing the same power source or power supply, sharing PCB or other mechanical support, sharing the same processor or computer, sharing the same modem or transceiver, or sharing the same communication port. A sensor actuator unit may include one or more sensors, each with its associated signal conditioner and A / D (if required), and one or more actuators, each with its associated signal conditioner and D / A, if required. A sensor unit, an actuator unit, and a sensor / actuator unit are collectively referred to as ‘field units’.

[0086] Fig. 2 is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an embodiment incorporating a Modified Audio Feedback (MAF) module configured to apply a time delay or frequency shift to the detected voice signal and superimpose the modified signal with the cancelling signal for therapeutic applications, according to some embodiments of the present invention.

[0087] Fig. 2 shows an embodiment incorporating a modified-audio-feedback module 20. The MAF module may receive the detected self-voice (typically from sensor 12) and applies a controlled modification - such as a delay, frequency shift. Alternatively, a metronome pulse, that is known to assist persons who stutter, or noise or a specific set of frequencies to mask or subdue psychological or other perceived sound such as tinnitus. The modified signal may be summed in adder 21 with the cancellation output before emission by transducer 24. This allows simultaneous operation as a DAF or FAF device while cancelling the natural self-voice. Users therefore may hear only the processed voice without the disturbing internal resonance.

[0088] Possible embodiments of the different components of cancelling arrangement of Fig.1, and the cancelling and MAF in Fig. 2, are described further below.

[0089] In some embodiments, the air sensor 12, can be implemented in a variety of forms, including, a dynamic, piezoelectric, capacitive, an optical or any other microphone. One or more directional microphones and noise reduction schemes for the suppression of background and environmental noise can optionally and preferably used. These include: subtraction of background sound (with separate microphone or microphones for detection of voice an background); voice parameter identification for improved voice detection quality and signal-to-noise ratio, cueing sensors to limit the detection time window to instances of active speech, or machine-learning analysis of the voice. Low inherent latency in the microphone may be advantageous to minimize the overall latency of the cancellation signal.

[0090] In some embodiments, the bone 10, and soft tissue 11 sensors (also referred herein as “bone-conduction sensor” and “soft-tissue vibration sensor”), can be implemented in a variety of forms of contact and non-contact sensors, including, a bone conduction, piezoelectric, acceleration, capacitive or optical transducer-receiver. Typically, background and environmental sound is almost entirely reflected by tissue and bone due to the very large acoustic impedance mismatch between air and tissue or bone, therefore essentially no interference from background or environmental sound is expected in these sensors. Optionally and additionally more than one tissue or bone sensor may be used for improved voice pickup. Furthermore, the background-free signal from such sensors may be fused with the air sensor input to reduce the background or environmental sounds in the air-channel. Such noise reducing signal fusion technologies include variations of the backgroundsuppression methods described above. Low inherent latency in the microphone may be advantageous to minimize the overall latency of the cancellation signal.

[0091] In some embodiments, the earphone transmitter 24 can be implemented in a variety of forms including an dynamic, piezoelectric, capacitive, or mechanical earphone / speaker. The soft tissue 28 and bone 26 transducers-transmitters can be implemented in a variety of forms of contact transmitters, including, a bone conduction, a piezoelectric, an acceleration, or a mechanical contact transducer-transmitter. Optionally more than one transducertransmitter for each channel may be used for improved performance. Some or all may be located inside the ear canal, or, additionally and alternatively, may be placed over the ear (in the case of the air channel) or near the ear, in the case of soft-tissue or bone transducertransmitters.

[0092] In some embodiments, the detection sensors of the bone and tissue channels are similar and, unless special measures are considered (for example mounting the bone sensor directly on bone: a tooth or using a skull implant), their location in contact with soft-tissue close to bone, is similar. It may therefore be sufficient to use only two pickup sensors - the air channel, 12, and the bone conduction sensor 10. The additional information obtainable from the soft-tissue sensor 11 may well be of lesser importance to the ability of the proposed system to effectively cancel the POV. Furthermore, it is possible to model the differences between the bone-borne and the soft-tissue channels and estimate the latter from the former. Taking this concept one further step, it is also possible to model the differences between the bone-borne and the soft-tissue borne signals from the airborne signal. In other words, both the bone-borne and the soft-tissue borne signal may be derived from the air-borne signal. In such cases, it would be sufficient to use only one sensor, the air sensor 12, and use this signal to define the spectral attenuation I gain and the spectral delay values to affect the cancellation of the POV in all three channels.

[0093] Fig. 3 is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating a configuration in which the system implements two cancelling channels — airborne and bone-conduction — while deriving soft-tissue cancellation from the bone-conduction path, according to some embodiments of the present invention.

[0094] In some embodiments, it may be possible to transmit the necessary soft-tissue cancelling signal through the bone. The bone cancelling signal may be then modified toaccommodate the effects of the soft-tissue channel. Fig. 3 shows this option with only two cancelling channels: air and bone. An adder 18 combines the sensed data from the three sensing channels (as shown in Fig. 3), or optionally, from one or two of them (not described in the Fig. 3) from which the cancelling bone-borne signal is generated.

[0095] Fig. 4 is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating a simplified configuration implementing only the airborne cancelling channel, wherein the cancelling signal is emitted through an air-conduction transducer positioned proximate to the ear, according to some embodiments of the present invention.

[0096] Taking the concept shown in Fig. 3 one further step, it is anticipated that, in some embodiments, both the bone-borne cancelling as well as the soft-tissue may be derived from the air-cancelling signal, so that only the air transmitter 24 is required to cancel the POV (Fig. 4).

[0097] In some embodiments, in one alternative POV cancelling arrangement, useful for hearing aid users, the cancelling airborne signal can be transferred into the hearing aid through its external interface, such as FM input, replacing the air-borne transducer transmitter in the system. This applies to a cancelling system which implements only the airborne cancelling channel, as described in Fig. 4. Of course, should additional cancellation channels be desired, these may be added to the basic hearing-aid instrumentation to address the more general case where two (as in Fig. 3) or three cancelling channels are deployed (as in Figs. 1 and 2). A second alternative may be applicable for people that have a cochlear implant. Here the cancelling acoustic airborne signal can be transferred to the cochlear implant rather than to the components of the ear. A third option may be to deploy a bone- anchored hearing system, in cases where such is available, to transmit a modified bone-borne cancelling signal, that has been modified, if necessary, to include components to cancel soft- tissue-bome signals. Naturally, in such cases there is no need to cancel the air-borne channel as it has had to have been non-functional to necessitate the bone-anchored hearing device in the first place.

[0098] The preceding discussion considers the devices and means to cancel POV and optionally add MAF in one ear. In general, the suggested system and device should be duplicated for application in both ears. It is possible that different signal parameters may be used in each ear to accommodate the different performance of each ear.

[0099] Figs. 1 through 4 show, conceptually, separate processing blocks. These can be implemented with analog or digital hardware modules, each performing one task. For example, filter modules with specified spectral attenuation, delay and phase can implement the filter blocks 13, 14, and 15 in Figs. 1- 4. Alternatively, these functions can be implemented with a digital platform, such as a processor, a digital signal processor (DSP), a field-programmable gate array (FPGA), application specific integrated circuit (ASIC), or other commensurate devices, all of which can be programmed to implement the processing and computational functions described above. Implementation with such a processor would also, naturally, require the incorporation of non-volatile memory storage to store the program and its parameters, suitable amplifiers and power supplies and a connector for programming and updating the software or firmware. Alternatively, a processing device, such as a personal computer or a smart phone for performing the analysis and processing functions.

[0100] Naturally, the use of any of the digital processors above also warrants application of advanced data and signal processing algorithms. Here, there is preference to deploy algorithms to identify specific speech parameters, such as pitch and phonemes. Such an algorithm is expected to perform better with cancellation of speech data. Such algorithms can use speech identification tools such as Hidden Markov models, Dynamic time warping (DTW)-based speech recognition, Neural networks, Deep feedforward and recurrent neural networks.

[0101] Figs. 5 A through 5C show schematic block diagrams of the processing stages in off-line experiments (or configurations) for determining the three transfer functions of the mouth-ear channel (speaker simulating the mouth 56, air gap and the microphone simulating the ear 52), the cancelling channel (canceling speaker 54 and microphone simulating the ear 52), and the mouth channel (mouth speaker 56 and mouth microphone 55). The cancelling signal, AC, to be used in the cancelling process is derived by these three transfer functions as detailed hereinafter. The figures show the stages in the procedure, where in each stage, the components that are inactive are marked with a cross sign.

[0102] Fig. 5 A is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an offline calibration arrangement for determining the mouth-to-ear transfer function by transmitting a reference signal through a mouth speaker and detecting the response at an ear microphone, according to some embodiments of the present invention.

[0103] Fig. 5 A shows the calibration process of the mouth-ear channel, in some embodiments. A calibration signal A’, stored on a personal computer, PC 69, is fed to the mouth speaker 56, and received by the microphone representing the ear 52, amplified 61, and recorded on the PC 69. This signal, referred to as A ’ME (A’ signal from Mouth to Ear), may be modified due to effects in the mouth speaker 56, the ear microphone 52, as well as any acoustic environmental effects such as delay, attenuation, reflections and reverberations. An inverse signal, -A ME, with the required amplitudes and phase of each frequency component is required for cancelling 4 ME.

[0104] Fig. 5B is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an offline calibration arrangement for determining the cancelling-channel transfer function by transmitting a reference signal through a cancellation speaker and detecting the response at the ear microphone, according to some embodiments of the present invention.

[0105] In some embodiments, to determine the transfer function of the cancelling channel (Fig. 5B), a suitable electronic audio calibration signal A’, stored on the PC 69, may be amplified 62, transmitted by cancelling speaker 54 to ear microphone 52, and optionally amplified 61. The resulting signal A’CE (A’ signal from Cancelling speaker to Ear), may thus incorporate any signal modifications within the cancelling channel. Here, the driving and receiving electronic circuits mentioned above may include amplifiers, filters and, in digital systems, a D / A before the speaker and an A / D after the microphone, as well as any acoustic effects of the acoustic environment, including delay, attenuation, reflections and reverberations. A’ may be designed in any format that incorporates all the relevant frequencies of the anticipated voice signal. One possible form of A’ can be a sinusoid with slowly varying frequency (often referred to as a “chirp”). Alternatively, A’ can be implemented with a recorded or synthetic voice.

[0106] Fig. 5C is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an offline calibration arrangement for determining the mouth-channel transfer function by transmitting a reference signal through the mouth speaker and detecting the response at a mouth microphone, according to some embodiments of the present invention.

[0107] Similarly, in some embodiments, the transfer function of the mouth channel (Fig. 5C), may be determined with a suitable electronic audio calibration signal A’, stored on thePC 69, transmitted by the mouth channel (speaker 56 to microphone 55, optionally amplified 64, and stored on PC 69). The resulting signal A’MM ’ signal from Mouth speaker to Mouth microphone), may incorporate any signal modifications due to the electronic components and environmental effects, as detailed above in relation to the other channels.

[0108] The transfer function (TF) that creates the cancelling signal, AC, can then be calculated by determining the transfer functions that may generate A ’ME, A ’CE and A ’MM. By way of example, the following describes a procedure in the frequency domain. Several other methods to determine the transfer and inverse transfer functions of a channel or system that can be applied here, including, convolution in the time-domain, direct learning approach, indirect learning approach, and various filter-based solutions, as known in the art.

[0109] In the following description, the notations FT{ } and FT-1{ } represent the Fourier Transform and the inverse Fourier Transform, respectively. The symbol ° represents the Hadamard product defined as (A°B)jj = A^By. The symbol 0 represents the Hadamard quotient with a complementary definition. The transfer function of the mouth to ear channel, TFME, may be calculated from FT{AME} and the Fourier Transform of the reference signal, FT{A} as TFME= FT{A'ME] 0 FT{A'}. Similarly, the Fourier Transform of the transfer function of the cancelling channel may be calculated as TFCE= FT{A'CE] 0 FT{A'}, and the Fourier Transform of the transfer function of the mouth channel may be calculated as TFCE= FT{A'MM] 0 FT{A'}. The Fourier Transform of the transfer function that generates the cancelling signal may be calculated as TF = - TFME0 ( TFMMO TFCE). TF may be then used to generate the cancelling signal, AC, to be fed to the cancelling speaker for cancelling the audio signal, AME, that is: AC = FT-1{TF O FT{AMM]}, where AMMis signal A after going through mouth speaker and mouth microphone.

[0110] Fig. 5D is a schematic block diagram of a system or device for cancelling a perception of a user’s own voice, illustrating an arrangement for demonstrating cancellation of airborne voice transmission by simultaneously transmitting an audio signal through the mouth speaker and a cancelling signal through the cancellation speaker, according to some embodiments of the present invention.

[0111] Fig. 5D shows schematically the arrangement for demonstrating offline the cancellation process, in some embodiments, where both the original voice signal, A, and its cancelling counterpart, AC, are stored on a PC 69, and generated from it. Both are optionallyamplified (amplifiers 63 and 62), then transmitted to the mouth speaker 56 and cancelling speaker 54, respectively. Both signals are detected by the ear microphone 52, optionally amplified to generate the combination of AME and -AME, which cancel each other. Any residual signal detected with the ear microphone due to imperfect cancelling can be recorded in the PC 69. We note here that the cancellation of the signals described above is considered to take place in the electronic form of the two signals AME and -AME. In practice, the signals may already cancel each other in their acoustic form, at the input to the microphone, before they are converted into their corresponding electronic form. The acoustic form of the signals before they are detected by the microphone cannot be analyzed in the described configuration, as only their electronic form can be measured. Therefore, the design and calibration necessarily consider the cancellation of the signal in electronic form. As, in the described embodiment, the two acoustic signals input into the microphone are transformed into electronic signals by identical functions; if the electronic signals are designed to cancel each other, the acoustic signals necessarily cancel as well.

[0112] It shall be understood that the transfer function, T F, that may be used to generate the cancelling signal, AC, from the audio signal to be cancelled, A, may, due to the calibration procedure described above, compensate for all the effects of the speakers, 54, 56, the microphones, 52, 55, the electronic circuits, the air gaps and any acoustic environmental effects, including delay, attenuation, reflections and reverberations. As long as the experimental setup and its environment remain unaltered, this transfer function, TF, may remain unaltered and can be used to calculate the required cancelling signal, AC, to be transmitted in the cancelling channel. It shall be noted that the above procedure may work efficiently as long as the system is linear and there are no zeroes in the transfer function. The former requirement is practically ensured if the signals are small. The latter difficulty is addressable by suitable processes known in the art. If some nondinear effects are present, these may be addressed by iterating the calculation of the transfer function, TF, to iteratively find a closer approximation to the required transfer function form. Fig. 5D does not detail this iterative process. The correction of the TF according to the residual signal is discussed further below.

[0113] Correcting TF according to the residual signal. First, signals and transfer functions should be defined (names of function or signal in small letter - means it is in thetime domain; names of function or signal in CAPITAL letter - means it is in the Frequency domain; multiplication and division in frequency domain are Hadamard operators).EM - Ear Mic Transfer Function [FREQ] r, R - Residual signal as measured after EM [time, FREQ]Rbm = R / EM Rbm is R before mic (acoustic signal in the ear)ES - Ear Speaker Transfer Function [FREQ] c, C - Canceling signal before the ES [time, FREQ]Cas= C -ES Cas is C after speaker (acoustic signal in the ear)MM - Mouth Mic Transfer Function [FREQ]MS - Mouth Speaker Transfer Function [FREQ] s, S - Signal played by MS and received by MM [time, FREQ]TF - The Transfer function that creates C from S (STF = C)

[0114] During system operation, it is possible to measure the power of r. If it exceeds a specified threshold, the processor may be configured to initiate a calculation to correct TF. The corrected TF should create a cancelling signal, c, that is configured to result in a zero, or at least smaller residual signal, r.

[0115] The processor may be configured to define:Co , Co - Canceling signal before the correction (TFo = TF before the correction);Ci , Ci - Canceling signal after the correction (TFi = TF after the correction)

[0116] The processor may be further configured to calculate:Output Cn Output CiInput S Input S

[0117] The processor may be configured to calculate the following conditions, which should be met to make the correction:Cas,l=Cas,0 ~ rbmC ES = Co'ES -R / EMCi = Co -R / (ES-EM)

[0118] In some embodiments, it may be beneficial to use two separate devices for the transmission and the detection of the audio signals, that is to transmit the reference signal A ’ by one device, such as the PC shown in Figs. 5A through 5D, and receive the A ’ME (or A ’CE) signal by another, such as a signal recorder, an oscilloscope or a spectrum analyzer (not depicted in Figs. 5A through 5D). In such situations the recorded signal A ’ and the recorded signal A ’ME may be synchronized to calculate TFM, similarly, the recorded signals A ’ and A ’CE may be synchronized to calculate TFc. Such synchronization may also be required in some situations where the signals above are transmitted and detected by the same device, such as the PC in Figs. 5 A through 5D.

[0119] To facilitate the required synchronization described above, a short synchronization signal can optionally be added to the calibration signal, A ’, such that it will remain apparent in the output signal, A ’ME or A ’CE. Such a synchronization signal can, in some embodiments, amongst other possibilities, take the form of a short high-frequency pulse, a burst of a few cycles at a certain frequency, a chirp of varying frequency, or short burst of cycles incorporating a phase inversion. On comparing^ ’ and the recording of A ’ME (or A ’CE), any offset in the synchronization between the two can then be readily measured and corrected. Additionally, or alternatively, the synchronization can also be accomplished by scanning for the best cancellation of the input audio signal by the cancelling signal.

[0120] It will be understood by the person skilled in the art that the above experimental procedure can be replicated, with the required modifications, to demonstrate cancelling voice signals in the bone-borne channel, or in the tissue-borne channel. Additionally, cancellation of voice signals in any combination of air-borne, bone-borne or tissue-borne channels can be demonstrated with any combination of air-borne, bone-borne or tissue-borne cancelling signals.

[0121] In some embodiments, it is also possible to calculate TF with spoken voice instead of recording played by the mouth speaker. This may require a modified TF calculation: The voice signal, v, may be detected by the mouth microphone (M) (also may be referred herein as “voice input sensor”) and the ear microphone (E) (also referred herein as “residual signal correction sensor”). The detected voice signal, after transformation to the frequency domain may be referred to as VM and VE. The transfer function (TF) may then be calculated by: TF = - VE / (VM• TFEE) , where TFEE is the combined transfer function of the ear speaker and the ear microphone. The minus sign inverts the result as required for cancelling the voicesignal that comes from the mouth through the air to the ear. This calibration process may be performed offline prior to using the device, or after change in the environment that change, for example the reflections between the mouth and the ear. In real-time, the canceling signal in the time domain (cs) may be calculated by: cs = FT’1{VM• TF}, or by convolution in time domain, and then it is fed into the ear speaker. This process can be implemented with the appropriate changes to the bone and soft-tissue channels as well.

[0122] The above description procedure can be replicated, with the required modifications, to demonstrate cancelling voice signals in the bone-borne channel, or in the tissue-bom channel. Additionally, cancellation of voice signal bome in any combination of airborne, bone-bome or tissue bom channels can be demonstrated with any combination of airborne, bone-bome or tissue-bome cancelling signals.

[0123] In some embodiments, the timing of the real-time cancellation process may require additional configurations. E.g., considering the cancellation of the air-channel voice, which, using a close-to-the-mouth input microphone 85, provides for a delay on the order of approximately 200 [is between the voice being spoken and the time it reached the ear. In modem processors, this delay may be sufficient for performing the processing required to generate an inverse of the heard voice. As for the sampling, however, this duration of 200 / zs, corresponds to the period of a 5 KHz signal. So, the sampling times for the portions of the voice signal that have frequency components of lower frequency will not be sampled for a sufficiently long duration to cover even one complete cycle. This might result in poor sampling of low frequency components, that may be insufficient to implement the anticipated processing to generate a cancelling signal. Nevertheless, it is possible to extend the sampling of lower frequency component by considering their presence at earlier times, then predicting their form in the time of relevance. Similar processing is known in the state of the art.

[0124] Extending the above to cancellation of the bone-bome and the tissue bom channels, the signals in these channels have even shorter delay and would require even shorter sampling durations. So, here too, using past information and predicting its current form may be required.

[0125] To avoid misconceptions we consider the underlying differences between the present invention and the well-known active noise control (ANC) devices, also known as noise cancellation: methods for reducing unwanted sound by the addition of a secondsound specifically designed to cancel the first. Systems use ANC to reduce background noise in many fields, including in pilot’s headsets, car telephone systems, industrial machinery, primarily air-condition systems in user environments and many others. Although the current invention is based on a similar physical phenomenon, namely the reduction or cancelling of an acoustic signal by a complementary anti-signal, the current invention provides completely different hardware arrangement, its implementation targets three noise tracks and its application is aimed at eliminating the effects of a PSH of POV. More specific differences are listed below.

[0126] ANC cancels surrounding, environmental or background noise, while the present solution cancels PSH of POV. Typically, ANC systems are limited to cancellation of deterministic, time-invariant noise and are unable to cancel transient signals. The present solution cancels transient speech signals. Typically, ANC systems are non-synchronous. The suggested solution is necessarily synchronous. Furthermore, it performs a spectral attenuation or gain on each spectral component and introduces a spectral delay of each spectral component. ANC addresses noise sources at unknown and variable orientations and distances, while the present solution addresses a determined position and distance of the voice source. Typically, ANC uses an omnidirectional microphone to pickup noise from the surroundings, while the present solution, in some embodiments, uses a directional microphone to pick up voice signals. ANC detect air-borne sound, while the present solution may be configured for detecting air, bone or tissue borne signals. ANC cancelling signal is typically limited to frequencies under IKHz, while the present invention is configured so as to cancel speech signal which includes substantial components up to at least 3.0 KHz. ANC cancels noise from external sources, effecting a single, undistorted interruption channel, while the present solution is configured so as to cancel the PSH of POV as it reaches the ear after travelling three acoustic channels - through air, through the skull bones and through the skull’s soft tissues. The present solution may operate with minimal latency, typically below 1ms. This requires selection of microphones and transducers with minimal latency, and preferably a microphone mounted near to the mouth to reduce the air-path delay. ANC considers only external sounds as they reach the device and there are no considerations of path delays; latency is typically less of an issue as ANC systems typically deal with deterministic noise. The present solution may incorporate speech-related software analysis to identify the important speech sounds for improved cancelling. ANC deals with moregeneral sound signals and in itself cannot benefit from analysis of the noise in terms of voice parameters.

[0127] Figs. 6A through 6D schematically illustrate the flow of four processes in the proposed arrangements for calibration of the system and cancellation of the POV. Figs. 6A and 6B show the flow of processes for the calibration of the mouth channel and the ear channel.

[0128] Fig. 6A is a flow chart illustrating a calibration process for determining the transfer function of the mouth speaker to the mouth microphone (MS-MM channel), including Fourier transformation of detected signals, according to some embodiments of the present invention.

[0129] Fig. 6 A shows the calibration process for the mouth speaker (MS or ML) to the mouth microphone (MM) which is the mouth channel. Here a digital audio reference signal, rs, that is stored in the system processor’s memory may be played (step 101), converted to an analog form (step 102) and transmitted as an acoustic signal with the MS (step 103). This reference signal, rs, may be broad band containing frequency components that span the full band of human hearing, ideally between 20 and 20,000 Hz, but practically covering a smaller band, such as 100 to 7,000 Hz or any other band used in practice by the proposed system. This process calibrates the characteristics of MS-MM channel, including any environmental echoes or air-transmission delays. The acoustic signal detected by the MM (step 104) may be converted to electronic form, rsi’. This signal may be modified by the combined characteristics of MS and MM. In the next stage this signal is digitized and Fourier transformed (step 105) to generate the representation of rsi’ in the frequency domain, RSf . Using a pre-calculated and stored Fourier transform of the reference audio signal, RS, the transfer function for the MS-MM combination, TFMS-MM, is calculated (step 106). This is indicated schematically by a Hadamard division in the frequency domain:_ RSi TFMS-MM_

[0130] Nevertheless, in practice a variety of algorithms to determine the transfer function can be used for example convolution in the time-domain, several filter types, and others. The transfer function, TFMS-MM, may then be stored (step 107) in the system’s memory for later use.

[0131] Fig. 6B is a flow chart illustrating a calibration process for determining the transfer function of the ear speaker to the ear microphone (EE-EM channel), including storage of the calculated transfer function for later use, according to some embodiments of the present invention.

[0132] Fig. 6B shows the calibration process for the ear earphone (EE in Fig. 6B) to the ear microphone (EM in Fig. 9B) channel. Here, a digital audio reference signal, rs, that may be stored in the system memory may be played (step 111), converted to an analog form (step 112) and transmitted as an acoustic signal with the EE (step 113). This reference signal, rs, may be broad band as indicated in the previous paragraph. This process is intended to calibrate the characteristics of EE-EM channel, including any environmental echoes or airtransmission delays. The acoustic signal detected by the EM (step 114) may be converted to electronic form, rs2’ . This signal may be modified by the combined characteristics of EE and EM. In the next stage this signal may be digitized and Fourier transformed (step 115) to generate the representation of rsi’ in the frequency domain, RSf . Using a pre-calculated and stored Fourier transform of the reference audio signal, RS, the transfer function for the EE- EM combination, TFEE.EM, may be calculated (step 116). This is indicated schematically by a Hadamard division in the frequency domain:

[0133] Nevertheless, in some embodiments, a variety of algorithms to determine the transfer function can be used for example convolution in the time-domain, several filter types, and others. The final transfer function, TFEE.EM, may then be stored (step 117) in the system’s memory for later use.

[0134] Fig. 6C is a flow chart illustrating a calibration process for determining the transfer function of the mouth speaker to the ear microphone (MS-EM channel) and combining previously calculated transfer functions to compute the overall transfer function of the system, according to some embodiments of the present invention.

[0135] Fig. 6C shows the calibration process for the mouth speaker, MS (or ML), to the ear microphone, EM. The process is akin to processes described above with the appropriate replacement of the devices being calibrated. Specifically, a digital audio reference signal, rs, may be played (step 121), converted to an analog form (step 122) and transmitted as an acoustic signal with the MS (step 123). The ear microphone EM may detect the acousticsignal (step 124) which may then be converted to electronic form, rs?’, which may be digitized and Fourier transformed (step 125) to generate the representation of rsi’ in the frequency domain, RS3’. Using a pre-calculated and stored Fourier transform of the reference audio signal, RS, the transfer function for the MS-EM combination, TFMS.EM, may be calculated (step 126) and stored the final transfer function, TFMS.EM. This transfer function may be used to calculate, together with the previous two transfer functions determined as described above, TFMS-MM, TFEE.EM, the overall transfer function of the system (in the Frequency domain, with Hadamard multiplication and division):_ TFMS.EMTF = - -TFMS-MMXTFEE.EM

[0136] In some embodiments, a variety of algorithms to determine the transfer function can be used for example convolution in the time-domain, several filter types, and others. The transfer function, TF, may then be stored (step 117) in the system’s memory for later use.

[0137] Fig. 6D is a flow chart illustrating a process for measuring and compensating for delay and environmental reflections affecting the cancelling signal, including synchronization of the cancelling signal with the arrival of the POV at the ear, according to some embodiments of the present invention.

[0138] Fig. 6D shows the flow of processes for the measurement of the effect of distance and environmental reflections on the signals actually detected by the system, in some embodiments. The delay relates to the distance between the mouth microphone MM and ear speaker EM transmitting the cancelling signal. The transfer function, TF, may be calculated and stored in the calibration process described above in relation to Fig. 6C above.

[0139] In the POV cancellation process, shown in Fig. 6D, the POV (step 131), that is referred to as the audio signal at mouth, or v, may be detected by the mouth microphone MM (step 132), converted to electronic analog signal, v' , digitized and Fourier transformed (step 133) to V'. As explained above, the transfer function, TF may be calculated in a different way, when using the voice of the user instead of the recordings played by a mouth speaker. This TF is then used to calculate the frequency-domain cancelling signal, VC=V *TF (step 134). Using an inverse Fourier transform, the time-domain voice cancelling signal, vc, may be derived (step 135). This signal may be delayed if needed to be synchronized in the time-domain with the POV as it reaches the ear (step 136), vc' , D / A andtransmitted in the vicinity of the ear through the ear speaker ES (step 137), where the original POV and the synchronized vc' cancel each other.

[0140] Fig. 7 is a flow chart illustrating a real-time cancellation process using a precalibrated transfer function and monitoring of any residual signal at the ear, wherein the residual signal is used to modify and update the transfer function for improved cancellation, according to some embodiments of the present invention.

[0141] Fig. 7 shows the flow of processes for the cancellation of the POV in real time using a pre-calibrated transfer function, Z , as described above with the addition of detecting any residual, un-cancelled portions of the POV, res. This measured res is then used to modify and update the transfer function. As in the previous description this cancellation process initially uses the transfer function TF, calculated and stored in the measurement process described above in relation to Fig. 6C. As indicated above, TF can be used as long as the environment configuration is unchanged, and should be re-measured (re-configured) whenever such a change occurs. This can be initiated manually by the user or automatically following identification of a significant degrading in the cancelled POV signal as can be monitored by the ear microphone (also referred herein as “residual signal correction sensor”) that can be used to monitor the residual cancelled POV signal.

[0142] To understand the form of the residual signal, res, and the way it is used to modify and update the transfer function TF, the POV signal may be considered as detected by the MM microphone v It shall be understood that all the multiplications and the divisions in the frequency domain, are Hadamard operators. In anticipation of an un-cancelled portion of this signal, such a portion, 8, may be included in the detected signal. In the Fourier domain this signal may be denoted as V +A. Consequently, the first cancelling signal, in the Fourier domain, that is obtained is expressed asVC = (V+A) * TF .

[0143] The operation of this cancelling signal, in the time domain, vc, may be considered incomplete, leaving the residual signal, res. Expanding this expression, shows, that the residual signal in the frequency domain is simply RES = A x TF x TFES.EM. A modified transfer function may then be determined that, when multiplied in the frequency domain with the POV as detected by microphone MM, V', generated a modified and improved cancelling signal given by

[0144] The corrected TFcorrcan now be used to calculate an improved version of the cancelling signal, vc. This approach provides for an efficient cancellation if the transfer functions of the ear speaker and the ear microphone are linear over the entire spectrum of the signal. In the more general case the effect of the transfer function, TFES.EM. This transfer function can be inferred from manufacturer’ s data or measured similarly to the measurement of the other two transfer functions described in connections with Figs. 6A, 6B and 6C.

[0145] In the above description it was assumed that a pre-calibration for the transfer function was obtained and that the environmental effects such as distance and reflections are invariant over time. In some embodiments, it is possible to alleviate both of these assumptions. Regarding the pre-calibration, as long as the first transfer function generates a cancelling signal that is reasonably close to the POV signal to be cancelled, the measurement of the residual signal and the correction of the transfer function as described above, should correct the transfer function to generate a more effective cancelling signal, whether in one step or in a sequence of iterations. Similarly, should the working environment change, the same process of measuring the residual signal and correcting the transfer function is expected to correct for the change experienced.

[0146] The notations present in figures should be read as follows: MS - Mouth speaker for calibration, also referred herein as “configuration transducer”; EM - Ear microphone for calibration, also referred herein as “configuration sensor” or “residual signal correction sensor”; MM - Mouth microphone for cancellation, also referred herein as “voice input sensor”; EE - Ear speaker for cancellation, also referred herein as “cancellation transducer”.

[0147] Fig. 8A is a schematic illustration of a head-mounted sound module including an air-channel input microphone, an air-cancelling speaker, and a bone-conduction transducer, configured for placement over the ear, which may be integrated into a system or device for cancelling a perception of a user’ s own voice, according to some embodiments of the present invention.

[0148] Fig. 8B is a schematic illustration of a head-mounted configuration with two sound modules of a system or device for cancelling a perception of a user’s own voice, secured by a head strap, each module including components for air and bone-conduction cancellation, according to some embodiments of the present invention.

[0149] Figs. 8 A and 8B schematically illustrate hardware of a self-hearing cancellation device according to some embodiments. The presentation of the embodiment in a configuration with two voice sensors is provided, on the airborne channel (12 in Figs. 1 through 4), and two cancelling signal channels, the air- and the bone-borne channels (26 and 24 respectively in Figs. 1 through 4). Fig. 8A shows the inner side of the head-mounted sound module 90 (viewed from the face that is eventually closest to the skin of the head of a user). This module may include an air-cancelling speaker, 54 (24 in Figs. 1 through 4) over the ear, located approximately opposite the entrance to the ear canal, or, within the ear canal. In addition, the device may further include an air-channel input microphone 85 and a bone transducer-transmitter 87. The bone transmitter-transducer may be brought into contact with user’s skin when the device is worn. The air-channel input microphone 85 is shown on a boom that is located close to the user’s mouth. Alternatively, this microphone can be implemented within the contour of the device 90 which entails a processing algorithm with a lower latency as further described below. A calibrating speaker may optionally be incorporated in the microphone boom (not shown in Figs. 8A and 8B). Optionally and alternatively the input microphone itself can be used as a speaker utilizing the duality function of most microphones that can also serve as speakers. Optionally other sensor and cancellation channels may be implemented in similar configurations, including: one airchannel input sensor and one air-cancelling channel and two pickup channels, air- and bone- borne, and two cancelling channels, air- and bone-borne where an additional bonetransmitter should be incorporated into the system. This second bone-transmitter can be included within the head-mounted sound module 90, but may alternatively be mounted at a different location to minimize potential cross-talk between the sensing bone-transmitter and the cancelling bone transmitter. One possible location is on the head strap 95 used to mount the sound module 90, or optionally two such modules for each ear, onto the user’s head (see Fig. 8B). Additional sensors (not shown in Fig. 8A) may be incorporated into the sound module for various practical and performance benefits, including: a mechanical, or noncontact sensor that determines secure mounting of each sound module on the head (if the bone-transducer is either not present or using its signal to monitor acoustic coupling quality interferes with its operation); and, additional microphones for monitoring and potentially filtering environmental interference.

[0150] Fig. 8B is a possible embodiment of the system. It schematically illustrates the mounting over the ear of the sound module 90. Two sound modules may be used, secured in position with a head-strap 95 that also presses the two sound modules onto either ear against the side of the user’s head 97. Such pressure ensures good contact for the bone transducers-transmitters, which also incorporate a spring-loaded interface for comfortable but effective acoustic coupling onto the head. Each of the two sound modules can be calibrated independently and adjusted to the specific performance required by each ear. Optionally and alternatively only one module can be used.

[0151] In one preferred configuration a smart phone 96 may be used as the processor for the device. It is shown here connected with a cable to the head-mounted sound modules, but it may optionally and preferably be connected with a wireless channel, such as Bluetooth or Wi-Fi. Other configurations of the sound modules are suitable for other applications of the invention. For audiological measurements the preferred hardware would be to incorporate the sensors and cancelling channels into the audiological headset. In such embodiments, sensing channels may not be required as the transmitted sound is controlled by the testing instrumentation. Accordingly, only calibration may be necessary.

[0152] It is noted that, for the purpose of this embodiment, a variety of microphones can be implemented, in various places. Nevertheless, it may be advantageous to use high- performance microphones with low distortions over the required audio spectrum to reduce the non-linearity and thereby the requirement for correcting distortion, for example, monitor microphones and earphones offer such high performance.

[0153] In another embodiment, the time critical calculations may be done by dedicated hardware and / or software in the headset itself.

[0154] In another preferred embodiment, when using an NN, a mouth microphone may not be required. In this situation, the NN may be trained to recognize the voice of the user in the ear microphone (also referred herein as “residual signal correction sensor”) and cancel only this signal. In this embodiment, it might be necessary to use prediction of the voice of the user, or of another signals, in order to have enough time to overcome the hardware and software delay.

[0155] According to the concept of the present invention, several alternative and complementary signal-processing architectures may be employed to improve the cancellation of a person’s own voice (POV). These embodiments address differentperformance requirements such as latency, adaptability, and robustness under varying acoustic conditions.

[0156] In one embodiment, the cancellation signal may be generated by an adaptive filter operating in the time domain, such as a Finite Impulse Response (FIR) or Infinite Impulse Response (HR) filter. The system may employ an adaptive algorithm, preferably a variation of the Least Mean Squares (LMS) or Recursive Least Squares (RLS) algorithm.

[0157] In such embodiments, the filter coefficients may be updated on a sample-by- sample or block-by-block basis to minimize the residual signal detected by the ear microphone. This configuration minimizes processing latency and enables real-time cancellation of the user’s voice while compensating for the transfer function of the DAC, amplifiers, speakers, microphones, and the acoustic path, including reverberations and reflections.

[0158] In another embodiment, the system may employ a Neural Network (NN) configured to model the non-linear characteristics of the user’s voice and the acoustic system. The NN may comprise Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM) units, or one-dimensional Convolutional Neural Networks (1D-CNN) optimized for time-series prediction.

[0159] In such embodiments, the NN may receive the input signal (and optionally the error signal) and predicts the required anti-noise waveform. Adaptation to changing acoustic conditions (e.g., fit leakage, varying reflections) may be achieved by: (a) online fine-tuning: updating the weights of specific layers (e.g., the final fully connected layer) via backpropagation during operation; (b) context-aware switching: selecting among a plurality of pre-trained model states based on detected environmental acoustic parameters.

[0160] In yet another embodiment, the system may employ a cascaded architecture combining the predictive capability of a Neural Network with the rapid adaptability of a linear filter. A Neural Network (fixed or slowly adapting) may process the input signal to extract and predict the target voice waveform, effectively acting as a “Virtual Reference Generator.” This stage isolates the user’s voice from background noise and predicts waveform evolution. At the next stage, the output of the NN may be fed into a fast-adapting linear filter (e.g., a short-tap LMS filter). This filter may adjust amplitude and phase of the NN-generated prediction to match the instantaneous acoustic transfer function of the ear canal.

[0161] In yet another embodiment, the signal -processing chain may include: (a) A Non- Linear Predictive Module (e.g., a Neural Network) configured to forecast the waveform of the user’s voice over a prediction horizon equal to or greater than the system’s electromechanical latency, thereby compensating for the non-causal nature of the cancellation loop; (b) A Linear Adaptive Filter that may apply real-time amplitude and phase adjustments to generate the final anti-noise signal. The adaptive filter may continuously update its coefficients based on the residual error signal, performing fine-grain synchronization and compensating for instantaneous variations in the secondary acoustic path (e.g., fit leakage or transducer variance) that the predictive module cannot track.

[0162] This arrangement may effectively decouples the task of temporal forecasting (performed by the Neural Network) from the task of acoustic channel matching (performed by the Linear Adaptive Filter), thereby achieving robust cancellation under dynamic conditions.

[0163] Fig. 9 is a flow chart illustrating an implementation of the transfer function as a finite impulse response (FIR) filter in the time domain, including convolution-based generation of the cancelling signal, according to some embodiments of the present invention.

[0164] Fig. 9 shows an example of implementation of the TF as an FIR filter in the time domain. In this example, the signal from the mouth microphone may be used as an input (the array “in”) and the cancelling signal is the output (array “out”). This approach generates the cancelling signal as a convolution of the input signal and the impulse response of the system (derived by an inverse Fourier transform of the transfer function). To save time and lower the calculation delay, the system may calculate only one value of the convolution, which is the next output sample, and then, as a new input sample arrived, move the input window one sample ahead, and do the convolution again for the next output sample. This allows saving time and having a lower delay. In this example, the length of the signal is 1 second, the sampling window is 256 samples and the filter, which is the TF in time domain, has also 256 values.

[0165] In principle, if the situation is such that only the speaker’s voice is present, an analog electronics solution is possible. Here, the signal from the input microphone, is electronically inverted and fed into the cancelling speaker. Care is required to ensure that the latency matches the delay in the signal time-of-flight. This option may be useful only when background sound is not present, as any background sound may arrive to the user’s ear witha different delay than the user’s voice. One application where such analog electronic cancellation may be beneficial related to the cancelling of self-hearing in audio testing.

[0166] Regarding the issue of sampling delay, the implementation of real time processing, necessarily may limit the sampling duration to the delay time between the speech and its reaching the ear. Even if we consider a typical mouth to ear distance of 10 cm the delay is 300 ( s in air; it is much shorter in the bone, on the order of 50 ( s. In practice the sampling time can be increased to the extent of the stationarity of human voice signals which is typically considered to extend to 10 or 20 ms. At 10 ms the sampling window spectral resolution is limited to some 100 Hz which may not be sufficient in practice, for some applications higher spectral resolutions may be required. One possibility to improve on this basic performance is to introduce spectral shifts to the sampled data so as to interlace the spectral bins of the analysis thereby improving the spectral resolution. Alternatively, the calculation of the inverse cancelling signal can be performed with a convolution-based algorithm in the time domain. It is possible to use an overlapping convolution and compute only one point each time. Such an algorithm introduces no delay and it is therefore possible to generate the required cancelling signal at the same rate as the input voice signal is sampled. Conversely, in such instance it is required to delay the output cancelling signal to synchronize it with the audio signal to be cancelled.

[0167] There are other standard filters, in addition to convolution, that may be used for the real-time calculation of the cancelling signal. These include HR and FIR filter and a variety of machine learning approaches which show promising potential for implementation of a real-time cancelling signal generation method.

[0168] One possibility to overcome the limited spectral resolution due to the limited sampling window, is to change the sampling interval to match the dominant spectral components of the voice signal. This can be accomplished by changing the size of the sampling window, so that the dominant frequencies of the signal will fall on the frequency ’ s bins (frequencies that are multiples of 1 / SamplingWindow). It is possible also to adjust the sampling frequency to have a convenient number of samples in the sampling window that is a power of 2.

[0169] In an alternative approach the cancelling signal can be generated from the airborne voice with a neural network (NN). Here a training session is performed where a wide variety of voice signals is used as input and a cancelling signal is sought as the output. Oncetraining is completed the NN serves to generate the required cancelling signal. Testing with this approach showed good results alleviating the limitations on frequency resolution and short sampling durations.

[0170] The NN performance can be improved if the NN is trained to recognize the voice of a specific speaker. In such a case, the NN can distinguish between the voice of the user and voices of other persons or any other sound, and cancel only the voice of the user without canceling other voices and sounds.

[0171] The digital signal processing algorithms and the NN implementations, can be designed to forecast the cancelling signal a few milliseconds in the future. This might be important in order to overcome hardware and / or software delays and have more time for the creation of the cancelling signal.

[0172] The potential applications and use cases are described below.

[0173] The invention has wide applicability in both medical and consumer contexts. Each use case below corresponds to a distinct embodiment or operating mode but shares the same fundamental cancellation mechanism.

[0174] Speech Fluency and Therapy (DAF / FAF): The system can simultaneously cancel the natural self-voice and generate delayed or frequency-altered feedback for users who stutter or clutter. Because the unwanted internal component is suppressed, the modified voice remains clear, improving speech fluency without discomfort.

[0175] Hearing-Aid Occlusion Reduction: Integrated into hearing aids, the invention reduces the hollow or occluded perception of one’s own voice by actively cancelling or attenuating the user voice while maintaining normal amplification of external sounds and voices. The calibration routine automatically adapts to the acoustic coupling of each individual ear mold or dome.

[0176] Audiological Testing and Calibration: In clinical audiometry, the system can suppress the subject’s self-voice during speech or threshold tests to avoid lowering the ear gain by the user own voice. When testing hearing in one ear, it can be used instead of masking with noise, to cancel hearing in the second ear to prevent phantom hearing. It may also serve as a calibration tool for measuring mouth-to-ear transfer functions in research and fitting procedures.

[0177] Tinnitus and Sound Therapy: The system may deliver therapeutic signals that cancel the perceived tinnitus sound. Cancelling signals can be generated in some cases bydetecting the related vibrations of the eardrum using a microphone within the ear canal. In other instances, the user can adjust the cancelling frequency of combination of frequencies by trial and error.

[0178] Psychological and Neurological Rehabilitation: For individuals with speech anxiety, dysarthria, or Alzheimer’ s-related communication difficulties, the device can modulate or attenuate self-voice perception to restore comfortable speaking patterns. Adaptive parameters can be personalized to each user’s cognitive and auditory profile.

[0179] Consumer Communication and Multimedia Devices: The invention can be implemented in headphones, conferencing headsets, or augmented-reality (AR / VR) systems to provide a more natural sense of speaking. Users can talk for extended periods without the typical “in-head” resonance or fatigue. Selective neural -network training enables one user’s voice to be suppressed locally while preserving the voices of others.

[0180] Fig. 10 is a flow diagram depicting a method for cancelling a perception of a user’s own voice, according to some embodiments of the present invention.

[0181] The method steps shown in Fig. 10 may be implemented with hardware and software modules as explained above with reference to Figs. 1-4, 5A-5D, 6A-6D, 7, 8 A and 8B above.

[0182] As can be seen in the provided description, the present invention represents a system and method for cancelling a perception of a user’s own voice as perceived through airborne, bone, and soft-tissue conduction paths. The suggested solution may operate with minimal latency, adapt to individual anatomical and environmental conditions, and maintain transparency to external sounds. Furthermore, the solution incorporates calibration routines to accurately model the transmission characteristics of these paths. It also allows to employ adaptive or machine-learning techniques to refine cancellation performance in real time.

[0183] Unless explicitly stated, the method embodiments described herein are not constrained to a particular order or sequence. Furthermore, all formulas described herein are intended as examples only and other or different formulas may be used. Additionally, some of the described method embodiments or elements thereof may occur or be performed at the same point in time.

[0184] While certain features of the invention have been illustrated and described herein, many modifications, substitutions, changes, and equivalents may occur to those skilled inthe art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.

[0185] Various embodiments have been presented. Each of these embodiments may of course include features from other embodiments presented, and embodiments not specifically described may include various features described herein.

Claims

CLAIMS1. A system for cancelling a perception of a user’ s own voice (POV), comprising: a voice input sensor configured to detect a source voice signal generated by the user; a cancellation transducer positioned proximate to the user's ear; and a processing module configured to: receive the source voice signal; generate a cancellation signal by applying a transfer function (TF) to the source voice signal, wherein the transfer function models a combined acoustic transmission path of air, bone, and soft-tissue from the user's mouth to the user's ear; and drive the transducer with the cancellation signal to destructively interfere with the user’s own voice.

2. The system of claim 1, wherein the processing module is configured to estimate a bone-conduction component and a soft-tissue component of the user's voice based solely on the airborne source voice signal detected by the voice input sensor, and combine these estimated components into the cancellation signal.

3. The system according to any one of claims 1-2, wherein the processing module calculates the cancellation signal one sample at a time using a sliding input window, such that a processing latency of the system is lower than a time-of-flight of the user's voice from the mouth to the ear.

4. The system according to any one of claims 1-3, wherein the processing module comprises a Neural Network trained to generate the cancellation signal from an input sampling window that is shorter than a period of a lowest frequency component of the user's voice, thereby compensating for a short acoustic delay of the transmission path.

5. The system according to any one of claims 1-4, wherein the processing module is configured to forecast the cancellation signal for a future time instance to compensate for hardware and software latencies.

6. The system according to any one of claims 1-5, wherein the cancellation transducer comprises at least one of: an air-conduction speaker, a bone-conduction exciter, or a direct electrical interface to a hearing aid.

7. The system according to any one of claims 1-6, further comprising a monitoring microphone at the user's ear, wherein the processing module calibrates the transfer function (TF) using the user's natural spoken voice by calculating a ratio of a signal at the monitoring microphone (VE) to a signal at the voice input sensor (VM), normalized by a transfer function of the monitoring microphone and cancellation transducer (TFEE).

8. The system according to any one of claims 1-7, wherein the transfer function is derived from an offline calibration process comprising: measuring a mouth-to-ear response (TFME), measuring a mouth-to-mouth response (TFMM), and measuring a cancellation-hardware-to-ear response (TFCE), and calculating the transfer function as TF = - TFME / ( TFMM • TFCE ).

9. A method for real-time cancellation of a user's self-hearing, comprising: detecting a voice signal generated by the user; transmitting a cancellation signal to the user's ear based on a transfer function; monitoring a residual signal at the user's ear; and updating the transfer function in real-time by subtracting a correction factor, wherein said correction factor is derived from the residual signal normalized by a transfer function of an ear-speaker and ear-microphone path.

10. The method of claim 9, further comprising dynamically adjusting a sampling window size or sampling frequency to align frequency bins with dominant spectral components of the user's voice.

11. The method according to any one of claims 9-10, further comprising generating a Modified Audio Feedback (MAF) signal, comprising at least one of a time delay (DAF) or frequency shift (FAF), and superimposing the MAF signal with the cancellation signal to treat speech fluency disorders.

12. The method according to any one of claims 9-11, applied to a hearing aid, wherein the cancellation signal reduces an occlusion effect caused by the hearing aid blocking an ear canal of the user.

13. The method according to any one of claims 9-12, further comprising detecting eardrum vibrations associated with tinnitus and generating the cancellation signal to suppress a perceived tinnitus sound.

14. A non-transitory computer-readable medium comprising instructions to: detect an audio signal at an ear-canal microphone; process the audio signal using a Neural Network trained to identify a specific voice signature of a user; and generate a cancellation signal only when the specific voice signature is identified, thereby preserving environmental sounds.

15. The medium of claim 13, wherein the instructions generate the cancellation signal based solely on input from the ear-canal microphone without input from a mouth- proximate microphone.

16. A device for cancelling a user’s perception of a user’s own voice (POV), comprising: at least one voice input sensor configured to detect a source signal associated with the user’s voice; at least one cancellation transducer configured to emit an acoustic cancelling signal; and at least one non-transitory memory device storing modules of instruction code; and and at least one processor operatively coupled to the at least one memory device, the at least one voice input sensor, and the at least one cancellation transducer, and being configured to execute the modules of instruction code, whereupon execution of said modules of instruction code, the at least one processor is configured to: receive the source signal from the at least one voice input sensor; generate the acoustic cancelling signal by applying a pre-configured transfer function (TF) representing a function of: (a) the source signal detected by the at least onevoice input sensor, and (b) a target signal obtained after passing the source signal through an acoustic transmission path including one or more of: (i) an air channel; (ii) a bone channel through a bone tissue of the user; and (iii) a soft-tissue channel through a soft tissue of the user, said acoustic cancelling signal being configured to destructively interfere with the target signal; and drive the at least one cancellation transducer with the acoustic cancelling signal.

17. The device of claim 16, wherein the at least one voice input sensor comprises at least one of: (a) a microphone; (b) a bone-conduction sensor; and (c) a soft-tissue vibration sensor; and the at least one cancellation transducer comprises at least one of: (i) an airconduction speaker; (ii) a bone-conduction exciter; and (iii) a soft-tissue transducer.

18. The device according to any one of claims 16-17, wherein the at least one processor is further configured to modify the source signal by applying a time delay or a frequency shift thereto, and drive the at least one cancellation transducer with the acoustic cancelling signal while superimposing the modified source signal onto the acoustic cancelling signal.

19. The device according to any one of claims 16-18, wherein the pre-configured transfer function (TF) is derived from an offline calibration process comprising: computing a mouth-to-ear transfer function as a function of: (a) a first reference signal as transmitted through a configuration transducer positioned proximate to the at least one voice input sensor; and (b) a first detected signal as obtained by detecting the first reference signal by a configuration sensor positioned proximate to the at least one cancellation transducer; computing a cancelling-channel transfer function as a function of: (a) a second reference signal as transmitted through the at least one cancellation transducer positioned proximate to the configuration sensor; and (b) a second detected signal as obtained by detecting the second reference signal by the configuration sensor; and computing a mouth-channel transfer function as a function of: (a) a third reference signal as transmitted through the configuration transducer positioned proximate to the at least one voice input sensor; and (b) a third detected signal as obtained by detecting the third reference signal by the at least one voice input sensor; andcalculating the TF in the frequency domain as a function of the mouth-to-ear transfer function; the cancelling-channel transfer function and the mouth-channel transfer function.

20. The device of claim 19, wherein the TF is calculated in the frequency domain as an inverted Hadamard element-wise division of the mouth-to-ear transfer function by the product of the mouth-channel transfer function and the cancelling-channel transfer function.

21. The device according to any one of claims 16-20, wherein the at least one processor is configured to generate the acoustic cancelling signal by: transforming the source signal into a spectral representation in the frequency domain; applying the TF to the spectral representation to compute a cancelling signal spectrum, wherein each spectral component of the cancelling signal spectrum is amplitude- and phase-adjusted so as to compensate for delay and reflections of the acoustic transmission path; converting the cancelling signal spectrum into a time-domain by performing an inverse Fourier Transform, thereby obtaining the acoustic cancelling signal; and synchronizing the acoustic cancelling signal with an estimated arrival of the user’s own voice at the user’s ear to achieve destructive interference.

22. The device according to any one of claims 16-21, further comprising a residual signal correction sensor operatively coupled to said at least one processor and positioned in proximity to the at least one cancellation transducer sensor; and wherein said at least one processor is further configured to: receive an input signal from the residual signal correction sensor upon driving the at least one cancellation transducer with the acoustic cancelling signal, said input signal representing an un-cancelled portion of the user’s own voice; determine the residual signal in the frequency domain as a function of the input signal combined with the TF and an ear-channel transfer function pre-computed as a function of: (a) a fourth reference signal as transmitted through the at least one cancellation transducer; and (b) a fourth detected signal as obtained by detecting the fourth reference signal by the residual signal correction sensor;update the TF based on the determined residual signal, thereby generating a corrected transfer function; and further generating the acoustic cancelling signal by applying the corrected transfer function, so as to reduce the residual signal.

23. A method for cancelling a user’s perception of a user’s own voice (POV), by at least one processor, comprising: receiving a source signal associated with the user’s voice from at least one voice input sensor; generating an acoustic cancelling signal by applying a pre-configured transfer function (TF) representing a function of: (a) the source signal detected by the at least one voice input sensor, and (b) a target signal obtained after passing the source signal through an acoustic transmission path including one or more of: (i) an air channel; (ii) a bone channel through a bone tissue of the user; and (iii) a soft-tissue channel through a soft tissue of the user, said acoustic cancelling signal being configured to destructively interfere with the target signal; and driving at least one cancellation transducer with the acoustic cancelling signal.

Images