Patents by Inventor Moshe Haiut
Moshe Haiut has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Patent number: 11831344Abstract: A method for providing codewords, the method may include receiving by an input buffer, one or more chunks of data; calculating a location of relevant bits within one or more chunks of data; wherein the relevant bits comprise (a) variable length retrieval information used during a retrieval process of a first type codeword, or (b) a second type codeword; performing the retrieval process of the first type codeword and retrieving the first type codeword from a memory unit that stores only a fraction of a codebook, the codebook comprises first type codewords and second type codewords; determining whether the relevant bits comprises the second type codeword or not; and outputting the second type codeword or the first type codeword, based on the determination.Type: GrantFiled: January 21, 2020Date of Patent: November 28, 2023Inventors: Tomer Malach, Moshe Haiut
-
Patent number: 11238887Abstract: A processor, that may include at least one neural network that comprises at least one leaky spiking neuron; wherein the at least one leaky spiking neuron is configured to directly receive an input pulse density modulation (PDM) signal from a sensor; wherein the input PDM signal represents a detected signal that was detected by the sensor; and wherein the at least one neural network is configured to process the input PDM signal to provide an indication about the detected input signal.Type: GrantFiled: June 28, 2018Date of Patent: February 1, 2022Assignee: DSP GROUP LTD.Inventor: Moshe Haiut
-
Patent number: 11163531Abstract: A method and a MAC unit that may include accumulation unit and a multiplier. A accumulation unit that includes a first part, a second part and a third part. The first part may calculate a truncated sum. The second part may be configured to (a) receive, during each calculation cycle, a carry out of an add operation performed during a calculation cycle, (b) receive a sign bit of an intermediate product calculated during the calculation cycle; and (c) calculate, by the counter logic, a counter logic value, and (d) convert, after a start of a last calculation cycle of the calculation cycles, an output value of the counter logic to an intermediate value having a two's complement format. The third part may be configured to calculate an output value of the MAC unit based on the intermediate value and a truncated sum calculated by the first part of the accumulation unit.Type: GrantFiled: June 21, 2019Date of Patent: November 2, 2021Assignee: DSP GROUP LTD.Inventors: Moshe Haiut, Assaf Ganor
-
Patent number: 11087205Abstract: A neural network that may include multiple layers of neural cells; wherein a certain neural cell of a certain layer of neural cells may include a first plurality of one-bit inputs; an adder and leaky integrator unit; and an activation function circuit that has a one-bit output; wherein the first plurality of one-bit inputs are coupled to a first plurality of one-bit outputs of neural cells of a layer that precedes the certain layer; wherein the adder and leaky integration unit is configured to calculate a leaky integral of a weighted sum of a number of one-bit pulses that were received, during a time window, by the first plurality of one-bit inputs; and wherein the activation function circuit is configured to apply an activation function on the leaky integral to provide a one-bit output of the certain neural cell.Type: GrantFiled: January 23, 2018Date of Patent: August 10, 2021Assignee: DSP GROUP LTD.Inventor: Moshe Haiut
-
Publication number: 20200234125Abstract: A method for providing codewords, the method may include receiving by an input buffer, one or more chunks of data; calculating a location of relevant bits within one or more chunks of data; wherein the relevant bits comprise (a) variable length retrieval information used during a retrieval process of a first type codeword, or (b) a second type codeword; performing the retrieval process of the first type codeword and retrieving the first type codeword from a memory unit that stores only a fraction of a codebook, the codebook comprises first type codewords and second type codewords; determining whether the relevant bits comprises the second type codeword or not; and outputting the second type codeword or the first type codeword, based on the determination.Type: ApplicationFiled: January 21, 2020Publication date: July 23, 2020Inventors: TOMER MALACH, Moshe Haiut
-
Publication number: 20200034117Abstract: A method and a MAC unit that may include accumulation unit and a multiplier. A accumulation unit that includes a first part, a second part and a third part. The first part may calculate a truncated sum. The second part may be configured to (a) receive, during each calculation cycle, a carry out of an add operation performed during a calculation cycle, (b) receive a sign bit of an intermediate product calculated during the calculation cycle; and (c) calculate, by the counter logic, a counter logic value, and (d) convert, after a start of a last calculation cycle of the calculation cycles, an output value of the counter logic to an intermediate value having a two's complement format. The third part may be configured to calculate an output value of the MAC unit based on the intermediate value and a truncated sum calculated by the first part of the accumulation unit.Type: ApplicationFiled: June 21, 2019Publication date: January 30, 2020Inventors: MOSHE HAIUT, Assaf Ganor
-
Publication number: 20190013037Abstract: A processor, that may include at least one neural network that comprises at least one leaky spiking neuron; wherein the at least one leaky spiking neuron is configured to directly receive an input pulse density modulation (PDM) signal from a sensor; wherein the input PDM signal represents a detected signal that was detected by the sensor; and wherein the at least one neural network is configured to process the input PDM signal to provide an indication about the detected input signal.Type: ApplicationFiled: June 28, 2018Publication date: January 10, 2019Inventor: Moshe HAIUT
-
Publication number: 20180225564Abstract: A neural network that may include multiple layers of neural cells; wherein a certain neural cell of a certain layer of neural cells may include a first plurality of one-bit inputs; an adder and leaky integrator unit; and an activation function circuit that has a one-bit output; wherein the first plurality of one-bit inputs are coupled to a first plurality of one-bit outputs of neural cells of a layer that precedes the certain layer; wherein the adder and leaky integration unit is configured to calculate a leaky integral of a weighted sum of a number of one-bit pulses that were received, during a time window, by the first plurality of one-bit inputs; and wherein the activation function circuit is configured to apply an activation function on the leaky integral to provide a one-bit output of the certain neural cell.Type: ApplicationFiled: January 23, 2018Publication date: August 9, 2018Inventor: Moshe Haiut
-
Patent number: 9859791Abstract: A method for charge-reuse, the method may include performing multiple repetitions of the steps of: operating a second capacitive load while the second capacitive load is disconnected from a first capacitive load; wherein the second capacitive load is a Microelectromechanical systems (MEMS) capacitive load or a Nanoelectromechanical systems (NEMS) capacitive load; electrically coupling a first capacitive load to a second capacitive load via a path that comprises an inductor; charging the first capacitive load with a second charge provided from the second capacitive load; electrically disconnecting the first capacitive load, the second capacitive load and the inductor from each other; feeding the inductor with a supply current provided by a supply circuit; disconnecting the inductor from the supply circuit and coupling the inductor to the first capacitive load; charging the first capacitive load by the inductor; electrically coupling the first capacitive load to the second capacitive load via the path that compType: GrantFiled: November 18, 2016Date of Patent: January 2, 2018Assignee: DSP GROUP LTD.Inventor: Moshe Haiut
-
Publication number: 20170222547Abstract: A method for charge-reuse, the method may include performing multiple repetitions of the steps of: operating a second capacitive load while the second capacitive load is disconnected from a first capacitive load; wherein the second capacitive load is a Microelectromechanical systems (MEMS) capacitive load or a Nanoelectromechanical systems (NEMS) capacitive load; electrically coupling a first capacitive load to a second capacitive load via a path that comprises an inductor; charging the first capacitive load with a second charge provided from the second capacitive load; electrically disconnecting the first capacitive load, the second capacitive load and the inductor from each other; feeding the inductor with a supply current provided by a supply circuit; disconnecting the inductor from the supply circuit and coupling the inductor to the first capacitive load; charging the first capacitive load by the inductor; electrically coupling the first capacitive load to the second capacitive load via the path that compType: ApplicationFiled: November 18, 2016Publication date: August 3, 2017Inventor: MOSHE HAIUT
-
Patent number: 9602920Abstract: A scheme is described to switch the power supply to the MEMS microphone on and off in a cyclic manner that is synchronized with the associated ADC sampling rate. In this way the MEMS microphone amplifier, whether it is a J-FET transistor or an operational amplifier, is off most of the cycle time, and is turned on only for a few micro-seconds prior to the sample-and-hold timing of the ADC device. By this method, the average power consumption of an existing analog MEMS microphone can be reduced by a factor of 10 or more.Type: GrantFiled: January 20, 2015Date of Patent: March 21, 2017Assignee: DSP GROUP LTD.Inventor: Moshe Haiut
-
Patent number: 9596534Abstract: Methods and systems are provided for controlling bone conduction, in which a bone conduction element may be used to output acoustic signals when it is in contact with a user. A bone conduction sensor may also be made in contact with the user, and used to obtain feedback relating to the outputting of the acoustic signals via the bone conduction element. The outputting of the acoustic signals may then be adaptively controlled based on processing of the feedback. The adaptive controlling may comprise adjusting components and/or functions related to or used in the outputting of the acoustic signals. For example, the adaptive controlling may comprise adjusting gain, frequency response, and/or equalization associated with a drive amplifier driving the bone conduction element.Type: GrantFiled: December 30, 2013Date of Patent: March 14, 2017Assignee: DSP Group Ltd.Inventors: Arie Heiman, Moshe Haiut, Uri Yehuday
-
Patent number: 9544788Abstract: A digital enhanced cordless telecommunication (DECT) base station, the DECT base station may include a transmitter that is arranged to transmit, at different points of time, downlink transmitted DECT signals to another DECT device that differs from the DECT base station; a receiver that is arranged to receive from the other DECT device downlink reception information indicative of attributes of downlink received DECT signals that were received by the other DECT device as a result of the transmission of the downlink transmitted DECT signals; a processor that is arranged to process the downlink reception information to detect a motion that affects a reception of the downlink received DECT signals by the other device.Type: GrantFiled: October 23, 2013Date of Patent: January 10, 2017Assignee: DSP Group Ltd.Inventors: Moshe Haiut, Lior Blanka
-
Publication number: 20160133255Abstract: A method for voice triggering, the method may include coupling, by an interface of a voice trigger sensor, the voice trigger sensor to a computer; receiving, by the voice trigger sensor, from the computer configuration information; configuring the voice trigger sensor by using the configuration information; coupling, by the interface, the voice trigger sensor to a target device during a voice activation period; receiving, by a processor of the voice trigger sensor, during the voice activation period, input signals; applying, by the processor, on the input signals a voice activation process to detect a voice command; and at least partially participating in an execution of the voice command.Type: ApplicationFiled: November 12, 2015Publication date: May 12, 2016Inventor: Moshe Haiut
-
Patent number: 9271076Abstract: Methods and systems are provided for enhanced stereo audio recordings in electronic devices. Stereophonic recording performance in an electronic device, using a first microphone and a second microphone in the electronic device, may be assessed; and processing of signals generated by the first microphone and the second microphone may be configured based on the assessed stereophonic recording performance. The configuring may comprises adaptively modifying the processing to enhance stereophonic recording performance, to match or approximate an ideal performance. The assessing of the stereophonic recording in the electronic device may be based on a type of each of the first microphone and the second microphone, and/or based on a spacing therebetween. The processing may be adaptively modified to simulate directional reception of signals by the first microphone and the second microphone when the microphones are omnidirectional.Type: GrantFiled: November 7, 2013Date of Patent: February 23, 2016Assignee: DSP Group LTD.Inventors: Arie Heiman, Moshe Haiut, Uri Yehuday
-
Patent number: 9154236Abstract: A system may be provided and may include a first transmitter; a first antenna, coupled to the first transmitter; wherein the first transmitter is arranged to transmit via the first antenna first radio frequency (RF) signals; a first receiver; a second antenna, coupled to the first receiver; the second antenna differs from the first antenna; an RF acquisition module that is arranged to acquire second RF signals that are representative of the first RF signals; and an RF suppression module, arranged to receive the second RF signals and to generate RF suppression signals for suppressing an effect of a transmission of the first RF signals on the first receiver, and to provide the RF suppression signals to the first receiver.Type: GrantFiled: September 24, 2012Date of Patent: October 6, 2015Assignee: DSP Group LTD.Inventors: Haim Kupershmidt, Kobi Ben Atar, Naftali Chayat, Moshe Haiut
-
Publication number: 20150215698Abstract: A scheme is described to switch the power supply to the MEMS microphone on and off in a cyclic manner that is synchronized with the associated ADC sampling rate. In this way the MEMS microphone amplifier, whether it is a J-FET transistor or an operational amplifier, is off most of the cycle time, and is turned on only for a few micro-seconds prior to the sample-and-hold timing of the ADC device. By this method, the average power consumption of an existing analog MEMS microphone can be reduced by a factor of 10 or more.Type: ApplicationFiled: January 20, 2015Publication date: July 30, 2015Inventor: Moshe Haiut
-
Patent number: 9077384Abstract: Methods and apparatus are described for reducing memory storage cells in a turbo decoder by storing only half the state metrics generated during a scan process. States associated with each bit transmission may be divided into couples and only one state from every state couple may be stored. In one example embodiment, only the state metric for a losing state of every state couple is saved, along with a single bit, e.g., 1 or 0, indicating whether the upper state or lower state of the state couple was the winner. The winning state may be reconstituted at a later stage. In this manner, for a code rate 1/3 and constraint length 3 turbo code, instead of storing 8*10=80 bits of state metrics for each systematic bit, only (4*10)+(4*1)=44 bits of scan state metrics data need be stored, a savings of nearly 50% regardless of the transistor technology used.Type: GrantFiled: October 15, 2012Date of Patent: July 7, 2015Assignee: MARVELL INTERNATIONAL LTD.Inventor: Moshe Haiut
-
Patent number: 9043211Abstract: In a mobile device, a bone conduction or vibration sensor is used to detect the user's speech and the resulting output is used as the source for a low power Voice Trigger (VT) circuit that can activate the Automatic Speech Recognition (ASR) of the host device. This invention is applicable to mobile devices such as wearable computers with head mounted displays, mobile phones and wireless headsets and headphones which use speech recognition for the entering of input commands and control. The speech sensor can be a bone conduction microphone used to detect sound vibrations in the skull, or a vibration sensor, used to detect sound pressure vibrations from the user's speech. This VT circuit can be independent of any audio components of the host device and can therefore be designed to consume ultra-low power. Hence, this VT circuit can be active when the host device is in a sleeping state and can be used to wake the host device on detection of speech from the user.Type: GrantFiled: May 8, 2014Date of Patent: May 26, 2015Assignee: DSP GROUP LTD.Inventors: Moshe Haiut, Arie Heiman, Uri Yehuday
-
Patent number: 8954814Abstract: Aspects of the disclosure can provide a method and an apparatus to decode a data stream based on multiple transmissions with efficient usages of storage and power resources. The method can include receiving a first plurality of encoded code blocks corresponding to a first transmission of a transport block, decoding the first plurality of encoded code blocks into decoded code blocks, error detecting the decoded code blocks, and storing a decoding history of the decoded code blocks. Further, the method can include receiving a second plurality of encoded code blocks corresponding to a retransmission of the transport block. The second plurality of encoded code blocks can map the first plurality of encoded code blocks, respectively. The method can selectively decode a subset of the second plurality of encoded code blocks based on the decoding history. In addition, the method can include storing soft bits for code blocks that failed decoding.Type: GrantFiled: January 18, 2013Date of Patent: February 10, 2015Assignee: Marvell World Trade Ltd.Inventors: Ronen Mayrench, Barak Ullman, Moshe Haiut, Shahar Fattal