Abstract: The invention is a multi-microphone voice processing SoC primarily for head worn applications. It bypasses the use of conventional pre-amp voice CODEC (ADC/DAC) chips all together by replacing their functionality with digital MEMS microphone(s) and digital speaker driver (DSD). Functionality necessary for speech recognition such as noise/echo cancellation, speech compression, speech feature extraction and lossless speech transmission are also integrated into the SoC. One embodiment is a noise cancellation chip for wired, battery powered headsets and earphones, as smart-phone accessory. Another embodiment is as a wireless Bluetooth noise cancellation companion chip. The invention can be used in headwear, eyewear glass, mobile wearable computing, heavy duty military, aviation and industrial headsets and other speech recognition applications in noisy environments.

Abstract: Systems and methods are described to time delay a signal output from an analog-to-digital converter (ADC). The ADC includes a digital sensor responsive to an analog field quantity. The digital sensor is configured to output an oversampled digital output signal at a sampling clock frequency. A time delay element is configured to receive as an input, the oversampled digital output signal and to output a time delayed oversampled digital output signal. A filter is configured to receive as an input the delayed oversampled digital output signal. The filter low pass filters and decimates to a lower sample rate the delayed oversampled digital output signal. An output includes a low pass filtered decimated delayed digital output signal, where the lower sample rate is less than the sampling clock frequency.

Abstract: A method of interfacing with a serving device from a wearable device worn by a user, the method includes establishing a lossless and wireless data link between the serving device and the wearable device. The method further includes collecting, by the wearable device, audio data from one or more microphones of the wearable device. The method also includes sending, by the wearable device, the collected audio data to the serving device through the lossless and wireless data link. The method may further include providing, by the serving device, speech recognition services associated with the audio data.

Abstract: Example embodiments include a method of reducing noise include forming a main signal and one or more reference signals at a beam-former based on at least two received audio signals, detecting voice activity at a voice activity detector, where the voice activity detector receives the main and reference signals and outputting a desired voice activity signal, adaptively cancelling noise at an adaptive noise canceller, where the adaptive noise canceller receives the main, reference, and desired voice activity signals and outputs an adaptive noise cancellation signal, and reducing noise at a noise reducer receiving the desired voice activity and adaptive noise cancellation signals and outputting a desired speech signal.

Abstract: Systems and apparatuses are described to automatically balance acoustic channel sensitivity. A computer readable medium containing executable computer program instructions, which when executed by an acoustic signal processing system, cause the acoustic signal processing system to perform a process that includes calculating a long-term power level of a main acoustic signal to obtain an averaged main acoustic signal. Segments of the main acoustic signal are excluded from the averaged main acoustic signal using a desired voice activity detection signal. A long-term power level of a reference acoustic signal is calculated to obtain, an averaged reference acoustic signal. Segments of the reference acoustic signal are excluded from the averaged reference acoustic signal using the desired voice activity detection signal. An amplitude correction signal is created using the averaged main acoustic signal and the averaged reference acoustic signal.

Abstract: Systems and methods are described to time delay a signal output from an analog-to-digital converter (ADC) The ADC includes a digital sensor responsive to an analog field quantity The digital sensor is configured to output an oversampled digital output signal at a sampling clock frequency. A time delay element is configured to receive as an input the oversampled digital output signal and to output a time delayed oversampled digital output signal. A filter is configured to receive as an input the delayed oversampled digital output signal. The fitter low pass filters and decimates tea lower sample rate the delayed oversampled digital output signal. An output includes a low pass filtered decimated delayed digital output signal, where the lower sample rate is less than the sampling clock frequency.

Abstract: Example embodiments include a method of reducing noise include forming a main signal and one or more reference signals at a beam-former based on at least two received audio signals, detecting voice activity at a voice activity detector, where the voice activity detector receives the main and reference signals and outputting a desired voice activity signal, adaptively cancelling noise at an adaptive noise canceller, where the adaptive noise canceller receives the main, reference, and desired voice activity signals and outputs an adaptive noise cancellation signal, and reducing noise at a noise reducer receiving the desired voice activity and adaptive noise cancellation signals and outputting a desired speech signal.

Abstract: Apparatuses and methods are described to identify desired audio. A first input of an apparatus is configured to receive a main signal. A second input of the apparatus is configured to receive a reference signal. A normalizer is configured to normalize a compressed main signal by a compressed reference signal to create a normalized main signal. A single channel normalized voice threshold comparator is configured to receive as an input the normalized main signal and to output a desired voice activity detection signal.

Abstract: An eyewear device includes a noise cancelling microphone array, and corresponding method for the same. The eyewear device and its method of use facilitates human-machine interaction using speech recognition by reducing unwanted noise that corrupts the desired speech and thereby reduces errors in speech. The eyewear device includes an eyeglasses frame, an array of microphones coupled to the eyeglasses frame, the array of microphones including at least a first microphone and a second microphone, the first microphone coupled to the eyeglasses frame about a temple region, the temple region can be located approximately between a top corner of a lens opening and a support arm, and providing a first audio channel output, and the second microphone coupled to the eyeglasses frame about an inner edge of the lens opening, and providing a second audio channel output. Embodiments can further include a third and fourth omni-directional microphone, providing a third and fourth audio channel, respectively.

Abstract: Systems and methods are described to reduce undesired audio. An adaptive noise cancellation unit receives a main signal and a reference signal. The main signal has a main signal-to-noise ratio; the reference signal has a reference signal-to-noise ratio. The reference signal-to-noise ratio is less than the main signal-to-noise-ratio. The adaptive noise cancellation unit reduces undesired audio from the main signal. An output signal from the adaptive noise cancellation unit is input to a single channel noise cancellation unit. The single channel noise cancellation unit further reduces undesired audio from the output signal to provide mostly desired audio. A filter control creates a control signal from the main signal and the reference signal to control filtering in the adaptive noise cancellation unit and to control filtering in the single channel noise cancellation unit.

Abstract: Systems, apparatuses, and methods are described to increase a signal-to-noise ratio difference between a main channel and reference channel. The increased signal-to-noise ratio difference is accomplished with an adaptive threshold for a desired voice activity detector (DVAD) and shaping filters. The DVAD includes averaging an output signal of a reference microphone channel to provide an estimated average background noise level. A threshold value is selected from a plurality of threshold values based on the estimated average background noise level. The threshold value is used to detect desired voice activity on a main microphone channel.

Abstract: Frequency domain signal extraction methods and apparatuses include receiving a reference signal, which contains mostly undesired audio and is substantially void of desired audio. The reference signal is decomposed into at least two reference frequency components. Filtering the at least two reference frequency components with at least two adaptive filters to form at least two filtered reference frequency components. The filtered reference frequency components are recombined in an IFFT component, to produce a filtered reference signal. A delayed signal is input to an adder. The delayed signal contains desired audio and undesired audio. The filtered reference signal is subtracted from the delayed signal to form an output signal containing desired audio. The output signal is decomposed into at least two frequency components. The filtering is adapted with the at least two frequency components.

Abstract: An acoustic system includes first one or mole acoustic elements designed and arranged in a first manner to facilitate generation of a first signal that includes mostly undesired audio, substantially void of desired audio, in response to a presence of the desired audio and the undesired audio. Second one or more acoustic elements are designed and arranged in a second complementary manner to facilitate generation of a second signal that includes both the desired and the undesired audio, in response to the presence of the desired audio and the undesired audio. A signal extraction component receives the first signal and the second signal. The signal extraction component further includes an inhibit component. The inhibit component is coupled to the first signal and the second signal. A delay element is coupled to a path of the second signal. The delay element introduces a deterministic delay to the second signal.

Abstract: A device and method to detect desired audio includes a ratio calculator. The ratio calculator calculates a ratio between a primary acoustic signal, and a reference acoustic signal. The primary acoustic signal contains desired audio and undesired audio and the reference acoustic signal contains mostly undesired audio, substantially void of undesired audio. A long-term mean value calculator is coupled to the ratio calculator. The long-term mean value calculator maintains an average of the ratio. A comparator is coupled to the ratio calculator and the long-term value calculator. The comparator compares the ratio with the average. Desired audio is detected when the ratio is greater than the average by a threshold amount.

Abstract: Apparatuses and methods are described to identify desired audio. A first input of an apparatus is configured to receive a main signal. A second input of the apparatus is configured to receive a reference signal. A normalizer is configured to normalize a compressed main signal by a compressed reference signal to create a normalized main signal. A single channel normalized voice threshold comparator is configured to receive as an input the normalized main signal and to output a desired voice activity detection signal.

Abstract: Systems and methods are described to extract desired audio from an apparatus to be worn on a user's wrist. The apparatus includes a wrist wearable device, configured to be worn on the user's wrist. The wrist wearable device includes a first microphone. The first microphone has a first response pattern. The first microphone is coupled to the wrist wearable device. The first microphone is positioned on the wrist wearable device to receive a voice signal from a user when the wrist wearable device is on the user's wrist.

Abstract: Systems and methods are described to extract desired audio from an apparatus to be worn on a user's head. The apparatus includes a head wearable device. A first microphone is coupled to the head wearable device, and is positioned on the head wearable device to receive a voice signal from the user when the head wearable device is on the user's head. A first signal from the first microphone is to be input as a main channel to a noise cancellation unit. A second microphone is coupled to the head wearable device. A first acoustic distance between the first microphone and the user's mouth is less than a second acoustic distance between the second microphone and the user's mouth when the head wearable device is on the user's head. A second signal from the second microphone is to be input as a reference channel to the noise cancellation unit. A first signal-to-noise ratio of the first signal from the first microphone is larger than a second signal-to-noise ratio of the second signal from the second microphone.

Abstract: Systems and methods are described to automatically balance acoustic channel sensitivity. A long-term power level of a main acoustic signal is calculated to obtain an averaged main acoustic signal. Segments of the main acoustic signal are excluded from the averaged main acoustic signal using a desired voice activity detection signal. A long-term power level of a reference acoustic signal is calculated to obtain an averaged reference acoustic signal. Segments of the reference acoustic signal are excluded from the averaged reference acoustic signal using a desired voice activity detection signal. An amplitude correction signal is created using the averaged main acoustic signal and the averaged reference acoustic signal.

Abstract: Systems and methods are described to create a desired voice activity detection signal. A main acoustic signal and a plurality of reference acoustic signals are compressed. The compressed main acoustic signal is normalized by the plurality of compressed reference acoustic signals to create a plurality of normalized compressed main acoustic signals. The plurality of normalized compressed main acoustic signals is processed with a plurality of single channel normalized voice threshold comparators to form a plurality of normalized desired voice activity detection signals. One of the plurality of normalized desired voice activity detection signals is selected from the plurality of normalized desired voice activity detection signals to output as the desired voice activity detection signal.

Abstract: A method of interfacing with a serving device from a wearable device worn by a user, the method includes establishing a lossless and wireless data link between the serving device and the wearable device. The method further includes collecting, by the wearable device, audio data from one or more microphones of the wearable device. The method also includes sending, by the wearable device, the collected audio data to the serving device through the lossless and wireless data link. The method may further include providing, by the serving device, speech recognition services associated with the audio data.