Abstract: Techniques are disclosed for a low-power and low-complexity speech onset detector (SOD) that uses a fractional-band filter structure and spectral subtraction technique to derive sub-band energy profiles to detect the onset of speech in the presence of noise. The SOD derives the sub-band energy profiles by filtering and down-sampling a full-band input audio signal using the fractional-bandwidth filter structure, which may be a low-pass filter with a cut-off frequency that is a fraction of the full bandwidth of the input signal. The SOD flexibly estimates the average noise energy across frames and the current frame speech energy in each sub-band to track noise and speech energy levels across the frames for each of the sub-bands to determine one or more band thresholds used to detect active speech. The sub-band energy profiles leverage any separation in frequency between noise and speech to detect the onset of speech in a target signal.

Abstract: Systems, methods, and devices detect wake signals included in audio signals. Methods include receiving a dataset including raw audio data, the raw audio data comprising a plurality of audio samples and associated metadata, and generating, using one or more processing elements, an augmented dataset based on the raw audio data, the augmented dataset comprising a plurality of annotations identifying types of raw audio data. Methods further include generating, using the one or more processing elements, a feature dataset by extracting features from the augmented dataset based, at least in part, on the plurality of annotations, and generating, using the one or more processing elements, a wake signal detection model based, at least in part, on the feature dataset, the wake signal detection model being a machine learning model trained based on the feature dataset.

Abstract: Described are techniques to recognize spoken wake word (WW) or command for human-machine interface using a speech recognition system that does not require any WW or command-matching speech data for training. The system uses the text or grapheme representation of the WW or commands for training before deployment. The technique includes receiving by the system a text representation of a target phrase of a target language. It includes training an acoustic model based on a speech database to distinguish speech signals according to acoustic units of the target language. The training of the acoustic model is independent of the target phrase, It includes constructing a recognition model based on the text representation of the target phrase and the acoustic model to recognize the target phrase in speech; and processing speech from a speaker based on the acoustic model and the recognition model to detect a presence of the target phrase.

Abstract: Described are techniques to recognize spoken wake word (WW) or command for human-machine interface using a speech recognition system that does not require any WW/command-matching speech data for training. The technique trains a tokenizer and includes breaking a word from a database into a plurality of combinations of unique sub-words by splitting the word at a plurality of different points for each of the combinations. The word includes one or more written units and one or more corresponding acoustic units. The technique maps the acoustic units constituting the word to the written units constituting the word to generate an acoustic-unit-to-written-unit mapping. The technique includes assigning a subset of the acoustic units to each of the unique sub-words based on the acoustic-unit-to-written-unit mapping to generate an acoustic-units-to-sub-words assignment for the word.

Abstract: Described are techniques to recognize spoken wake word (WW) or command for human-machine interface using a speech recognition system that does not require any WW or command-matching speech data for training. The system uses the text or grapheme representation of the WW or commands for training before deployment. The technique includes generating a matrix of statistics characterizing phonetic modeling of an acoustic model that distinguishes speech signals according to a plurality of acoustic units of a language. It further includes constructing a speech recognition model based on the matrix of statistics to recognize a target phrase in speech. It further includes processing an audio signal based on the acoustic model and the speech recognition model to detect a presence of the target phrase.

Abstract: Described are techniques to recognize spoken wake word (WW) or command for human-machine interface using a speech recognition system that does not require any WW/command-matching speech data for training. The system uses the text or grapheme representation of the WW or commands for training before deployment. The technique includes receiving a target phrase for recognition by a speech recognition model. The technique includes analyzing a sequence of acoustic units representative of the target phrase when the target phrase is spoken to generate offline analysis data. The technique further includes constructing the speech recognition model based on the offline analysis data to decode speech signals of the target phrase according to the acoustic units. The technique further includes processing speech based on the speech recognition model to detect a presence of the target phrase.

Abstract: Disclosed are methods and systems for a Bluetooth Low Energy (BLE) receiver to reduce the number of retransmission of packets needed to receive an error free packet so as to improve channel throughput. Techniques to reduce the number of retransmissions include a combination of processing of the header of the received packets to increase the number of corrupted packets available for reconstructing the original payload and bit error correction (BEC) of the payload of the corrupted packets. Header processing may include making available for payload reconstruction a packet whose received access address differs by no more than 1-bit from an assigned address of the receiver provided at least one of the corrupted packets used in the reconstruction contains an error-free access address. Header processing may also include using a prior error-free decoded length of the packet to aid in the determination of the length field of a current packet.

Abstract: Techniques are disclosed for a low-power and low-complexity speech onset detector (SOD) that uses a fractional-band filter structure and spectral subtraction technique to derive sub-band energy profiles to detect the onset of speech in the presence of noise. The SOD derives the sub-band energy profiles by filtering and down-sampling a full-band input audio signal using the fractional-bandwidth filter structure, which may be a low-pass filter with a cut-off frequency that is a fraction of the full bandwidth of the input signal. The SOD flexibly estimates the average noise energy across frames and the current frame speech energy in each sub-band to track noise and speech energy levels across the frames for each of the sub-bands to determine one or more band thresholds used to detect active speech. The sub-band energy profiles leverage any separation in frequency between noise and speech to detect the onset of speech in a target signal.

Abstract: Disclosed are methods and systems for a Bluetooth Low Energy (BLE) receiver to reduce the number of retransmission of packets needed to receive an error free packet. A softbit metric of a demodulator may be used to identify likely bit error positions of a corrupted packet and to correct bits of the corrupted packet corresponding to the identified bit error positions. When a demodulated packet fails the CRC, a receiver may identify one or more hypothesized bit error positions for the demodulated bits of the packet based on the softbit metric. The receiver may flip one or more of the demodulated bits corresponding to the hypothesized bit error positions. The receiver may determine if the packet after flipping the demodulated bits passes the CRC. If the CRC passes, the hypothesized bit error positions identifies the bit error positions of the corrupted packet and the receiver has corrected the bit errors.

Abstract: Systems, methods, and devices detect audio signals. Methods may include receiving an audio input at an audio front end circuit, starting, using a low power circuit, one or more buffers in response to receiving the audio input, and identifying, using the low power circuit, speech included in the audio input. Methods may also include identifying, using the low power circuit and a high performance circuit, a wake word based, at least in part, on the identified speech, the high performance circuit being configured to verify the wake word identified by the low power circuit.

Abstract: Disclosed are methods and systems for using softbit decoding techniques in retransmission-based networks for error concealment of packets corrupted by bit-errors. The softbit decoding techniques derive softbit information from multiple corrupted hardbits of the retransmitted packet to aid a softbit decoder in decoding the packet. The approach realizes improved error concealment capability while maintaining a simple system architecture. A retransmission softbit module is inserted between a channel decoder used for channel-decoding and demodulating a compressed packet and the softbit decoder. The retransmission softbit module may derive an accumulated softbit packet from multiple corrupted copies of the packet received from the channel decoder, make bit decisions based on the accumulated softbit packet, and derive reliability information for the bit decisions. The bit decisions may be a majority decision packet (MDP) created using a majority voting scheme.

Abstract: Implementations disclosed describe techniques and systems that improve wireless connectivity between devices that support multiple operation modes, by monitoring a quality of the established wireless connection and performing a change of the operation mode in response to changing conditions. The disclosed techniques include obtaining metrics characterizing quality of the wireless connection, identifying a utility function that is specific to an application supported by the wireless connection, computing multiple values of the utility function for various operation modes, based on the obtained metrics, and initiating a change of the operation mode based on a comparison of the computed utility values.

Abstract: Systems, methods, and devices seamlessly playback data files using one or more wireless connections. Methods include establishing a first wireless connection and a second wireless connection between a first device and a second device, the first wireless connection using a first communication protocol. Methods further include transmitting audio data via the first wireless connection and determining a switch should be initiated based on one or more signal quality metrics representing an estimate of a quality of the first wireless connection. Methods also include operating the second wireless connection in an active mode and switching to the second wireless connection using a second communication protocol, and the switching comprising beginning encapsulation of the audio data in Bluetooth packets for transmission in accordance with the second communication protocol, wherein the Bluetooth packets are encapsulated in one or more data packets of the second communication protocol.

Abstract: Techniques are disclosed for a low-power and low-complexity speech onset detector (SOD) that uses a fractional-band filter structure and spectral subtraction technique to derive sub-band energy profiles to detect the onset of speech in the presence of noise. The SOD derives the sub-band energy profiles by filtering and down-sampling a full-band input audio signal using the fractional-bandwidth filter structure, which may be a low-pass filter with a cut-off frequency that is a fraction of the full bandwidth of the input signal. The SOD flexibly estimates the average noise energy across frames and the current frame speech energy in each sub-band to track noise and speech energy levels across the frames for each of the sub-bands to determine one or more band thresholds used to detect active speech. The sub-band energy profiles leverage any separation in frequency between noise and speech to detect the onset of speech in a target signal.

Abstract: Disclosed are methods and systems for a Bluetooth Low Energy (BLE) receiver to reduce the number of retransmission of packets needed to receive an error free packet so as to improve channel throughput. Techniques to reduce the number of retransmissions include a combination of processing of the header of the received packets to increase the number of corrupted packets available for reconstructing the original payload and bit error correction (BEC) of the payload of the corrupted packets. Header processing may include making available for payload reconstruction a packet whose received access address differs by no more than 1-bit from an assigned address of the receiver provided at least one of the corrupted packets used in the reconstruction contains an error-free access address. Header processing may also include using a prior error-free decoded length of the packet to aid in the determination of the length field of a current packet.

Abstract: Disclosed are methods and systems for a Bluetooth Low Energy (BLE) receiver to reduce the number of retransmission of packets needed to receive an error free packet so as to improve channel throughput. Techniques to reduce the number of retransmissions include a combination of processing of the header of the received packets to increase the number of corrupted packets available for reconstructing the original payload and bit error correction (BEC) of the payload of the corrupted packets. Header processing may include making available for payload reconstruction a packet whose received access address differs by no more than 1-bit from an assigned address of the receiver provided at least one of the corrupted packets used in the reconstruction contains an error-free access address. Header processing may also include using a prior error-free decoded length of the packet to aid in the determination of the length field of a current packet.

Abstract: Systems, methods, and devices detect audio signals. Methods may include receiving an audio input at an audio front end circuit, starting, using a low power circuit, one or more buffers in response to receiving the audio input, and identifying, using the low power circuit, speech included in the audio input. Methods may also include identifying, using the low power circuit and a high performance circuit, a wake word based, at least in part, on the identified speech, the high performance circuit being configured to verify the wake word identified by the low power circuit.

Abstract: Systems, methods, and devices detect wake signals included in audio signals. Methods include receiving a dataset including raw audio data, the raw audio data comprising a plurality of audio samples and associated metadata, and generating, using one or more processing elements, an augmented dataset based on the raw audio data, the augmented dataset comprising a plurality of annotations identifying types of raw audio data. Methods further include generating, using the one or more processing elements, a feature dataset by extracting features from the augmented dataset based, at least in part, on the plurality of annotations, and generating, using the one or more processing elements, a wake signal detection model based, at least in part, on the feature dataset, the wake signal detection model being a machine learning model trained based on the feature dataset.

Abstract: Described herein are devices, methods, and systems for detecting a phrase from uttered speech. A processing device may determine a first model for phrase recognition based on a likelihood ratio using a set of training utterances. The set of utterances may be analyzed by the first model to determine a second model, the second model comprising a training state sequence for each of the set of training utterances, and wherein each training state sequence indicates a likely state for each time interval of a corresponding training utterance. A determination of whether a detected utterance corresponds to the phrase may be based on a concatenation of the first model and the second model.

Abstract: Systems, methods, and devices seamlessly playback data files using one or more wireless connections. Methods include establishing a first wireless connection between a source device and a sink device, the first wireless connection using a first transmission protocol and transmitting audio data via the first wireless connection. Methods further include determining a switch should be initiated based on one or more signal quality metrics, the one or more signal quality metrics representing, at least in part, an estimate of a quality of the first wireless connection. Methods also include switching to a second wireless connection between the source device and the sink device, the second wireless connection using a second transmission protocol, and the second wireless connection using data packets encapsulated for transmission in accordance with the second transmission protocol.