Patents by Inventor Alexander Kjeldaas
Alexander Kjeldaas 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: 10084921Abstract: Systems and methods are provided for handling concurrent speech in which temporally overlapping first speech data and second speech data is received from respective first and second participants of a session. A speech policy applied to the speech data specifies dropping the second speech when it interrupts the first speech within a first interval of the first speech data. The first interval is temporally bounded by the beginning of the first speech and a first predetermined amount of time after the beginning of the first speech. The speech policy specifies outputting the first speech data and then outputting the second speech data when the second speech data interrupts a second interval of the first speech data. The second interval of the first speech data is temporally bounded by the end of the first speech data and a second predetermined amount of time before the end of the first speech data.Type: GrantFiled: July 18, 2017Date of Patent: September 25, 2018Assignee: GOOGLE LLCInventors: Serge Lachapelle, Alexander Kjeldaas
-
Publication number: 20170318158Abstract: Systems and methods are provided for handling concurrent speech in which temporally overlapping first speech data and second speech data is received from respective first and second participants of a session. A speech policy applied to the speech data specifies dropping the second speech when it interrupts the first speech within a first interval of the first speech data. The first interval is temporally bounded by the beginning of the first speech and a first predetermined amount of time after the beginning of the first speech. The speech policy specifies outputting the first speech data and then outputting the second speech data when the second speech data interrupts a second interval of the first speech data. The second interval of the first speech data is temporally bounded by the end of the first speech data and a second predetermined amount of time before the end of the first speech data.Type: ApplicationFiled: July 18, 2017Publication date: November 2, 2017Inventors: Serge Lachapelle, Alexander Kjeldaas
-
Patent number: 9742921Abstract: Systems and methods are provided for handling concurrent speech in which first speech data is received from a first participant of a session and second speech data is received from a second participant of the session. The second speech data includes a pause. The second speech data temporally overlaps the first speech data. A determination is made as to whether the first speech data exceeds a predetermined length. When the first speech data exceeds the predetermined length, the first speech data is outputted and then the second speech data of the second participant is outputted without the pause. When the first speech data does not exceed the predetermined length, the first speech data is outputted and then the second speech data is outputted with the pause.Type: GrantFiled: October 27, 2016Date of Patent: August 22, 2017Assignee: GOOGLE INC.Inventors: Serge Lachapelle, Alexander Kjeldaas
-
Publication number: 20170048394Abstract: Systems and methods are provided for handling concurrent speech in which first speech data is received from a first participant of a session and second speech data is received from a second participant of the session. The second speech data includes a pause. The second speech data temporally overlaps the first speech data. A determination is made as to whether the first speech data exceeds a predetermined length. When the first speech data exceeds the predetermined length, the first speech data is outputted and then the second speech data of the second participant is outputted without the pause. When the first speech data does not exceed the predetermined length, the first speech data is outputted and then the second speech data is outputted with the pause.Type: ApplicationFiled: October 27, 2016Publication date: February 16, 2017Inventors: Serge Lachapelle, Alexander Kjeldaas
-
Patent number: 9491300Abstract: A system having one or more processors and a memory receives both speech data from first and second participants of a session. The system outputs the speech of the first participant. The system outputs the speech of the second participant concurrent with the speech of the first participant when the length of time of the speech data of the first participant is more than a predetermined threshold amount. The system outputs the speech data of the second participant in accordance with an adjustment of the speech of one or more participants of the session that includes delaying output of the speech data of the second participant until after the speech data of the first participant has been outputted when the length of time of the speech data of the first participant is less than the predetermined threshold amount.Type: GrantFiled: March 2, 2016Date of Patent: November 8, 2016Assignee: GOOGLE INC.Inventors: Serge Lachapelle, Alexander Kjeldaas
-
Publication number: 20160182728Abstract: A system having one or more processors and a memory receives both speech data from first and second participants of a session. The system outputs the speech of the first participant. The system outputs the speech of the second participant concurrent with the speech of the first participant when the length of time of the speech data of the first participant is more than a predetermined threshold amount. The system outputs the speech data of the second participant in accordance with an adjustment of the speech of one or more participants of the session that includes delaying output of the speech data of the second participant until after the speech data of the first participant has been outputted when the length of time of the speech data of the first participant is less than the predetermined threshold amount.Type: ApplicationFiled: March 2, 2016Publication date: June 23, 2016Inventors: SERGE LACHAPELLE, ALEXANDER KJELDAAS
-
Patent number: 9356965Abstract: A method performed at an electronic device with one or more processors and memory includes obtaining a public record including a plurality of commit nodes, where a respective commit node of the plurality of commit nodes includes: (i) a set of blobs corresponding to a collection of data; (ii) at least one cryptographic hash of one of the set of blobs; and (iii) at least one additional cryptographic hash of a parent node. The method includes: caching the public record; and obtaining, at a time after the obtaining, information corresponding to a new commit node, where the new commit node: includes a cryptographic hash of a parent node; and is associated with an updated collection of data. The method includes validating the new commit node; and in accordance with a determination that the new commit node is valid, updates the cached public record to include the new commit node.Type: GrantFiled: December 30, 2013Date of Patent: May 31, 2016Inventor: Alexander Kjeldaas
-
Patent number: 9313335Abstract: A system having one or more processors and a memory, receives both speech data from first and second participants of a session. The system outputs the speech of the first participant. The system outputs the speech of the second participant in accordance with an adjustment of the speech of a participant of the session when the speech of the second participant temporally overlaps less than a first predetermined threshold amount of a terminal portion of the speech of the first participant. The system drops the speech of the second participant when the speech of the second participant temporally overlaps more than the first predetermined threshold amount of the terminal portion of the speech of the first participant. Optionally, the system adjusts the speech of a participant of the session by delaying output of the speech of the second participant.Type: GrantFiled: September 13, 2013Date of Patent: April 12, 2016Assignee: GOOGLE INC.Inventors: Serge Lachapelle, Alexander Kjeldaas
-
Patent number: 9215458Abstract: A system, apparatus, and method for encoding a plurality of frames in a video stream with temporal scalability. The method includes identifying a non-uniform sequence of time values within a period, determining a frame corresponding to each time value in the non-uniform sequence, within at least one period, and assigning each of the determined frames to one of a plurality of temporal encoding layers.Type: GrantFiled: May 13, 2014Date of Patent: December 15, 2015Assignee: GOOGLE INC.Inventor: Alexander Kjeldaas
-
Patent number: 9210058Abstract: A system having one or more processors and a memory, sends a plurality of test audio packets at a level of signal complexity deviating from a model level of signal complexity to a destination device through one or more networks. The system then receives a response to the plurality of test audio packets, where the response is indicative of a value for a quality of service characteristic associated with the one or more networks, and where the value for the quality of service characteristic is determined by how the plurality of test audio packets deviate from the model level of signal complexity when received by a remote device. In response to receiving the response to the plurality of test audio packets, the system activates a signal correction action when the value for the quality of service characteristic fails to meet a performance threshold.Type: GrantFiled: March 8, 2013Date of Patent: December 8, 2015Assignee: GOOGLE INC.Inventors: Alexander Kjeldaas, Serge Lachapelle
-
Publication number: 20150188945Abstract: A method performed at an electronic device with one or more processors and memory includes obtaining a public record including a plurality of commit nodes, where a respective commit node of the plurality of commit nodes includes: (i) a set of blobs corresponding to a collection of data; (ii) at least one cryptographic hash of one of the set of blobs; and (iii) at least one additional cryptographic hash of a parent node. The method includes: caching the public record; and obtaining, at a time after the obtaining, information corresponding to a new commit node, where the new commit node: includes a cryptographic hash of a parent node; and is associated with an updated collection of data. The method includes validating the new commit node; and in accordance with a determination that the new commit node is valid, updates the cached public record to include the new commit node.Type: ApplicationFiled: December 30, 2013Publication date: July 2, 2015Inventor: Alexander Kjeldaas
-
Patent number: 8724712Abstract: A system, apparatus, and method for encoding a plurality of frames in a video stream with temporal scalability. The method includes identifying a non-uniform sequence of time values within a period, determining a frame corresponding to each time value in the non-uniform sequence, within at least one period, and assigning each of the determined frames to one of a plurality of temporal encoding layers.Type: GrantFiled: July 7, 2011Date of Patent: May 13, 2014Assignee: Google Inc.Inventor: Alexander Kjeldaas
-
Publication number: 20140078938Abstract: A system having one or more processors and a memory, receives both speech data from first and second participants of a session. The system outputs the speech of the first participant. The system outputs the speech of the second participant in accordance with an adjustment of the speech of a participant of the session when the speech of the second participant temporally overlaps less than a first predetermined threshold amount of a terminal portion of the speech of the first participant. The system drops the speech of the second participant when the speech of the second participant temporally overlaps more than the first predetermined threshold amount of the terminal portion of the speech of the first participant. Optionally, the system adjusts the speech of a participant of the session by delaying output of the speech of the second participant.Type: ApplicationFiled: September 13, 2013Publication date: March 20, 2014Applicant: Google Inc.Inventors: Serge Lachapelle, Alexander Kjeldaas
-
Publication number: 20140078916Abstract: A system having one or more processors and a memory, sends a plurality of test audio packets at a level of signal complexity deviating from a model level of signal complexity to a destination device through one or more networks. The system then receives a response to the plurality of test audio packets, where the response is indicative of a value for a quality of service characteristic associated with the one or more networks, and where the value for the quality of service characteristic is determined by how the plurality of test audio packets deviate from the model level of signal complexity when received by a remote device. In response to receiving the response to the plurality of test audio packets, the system activates a signal correction action when the value for the quality of service characteristic fails to meet a performance threshold.Type: ApplicationFiled: March 8, 2013Publication date: March 20, 2014Inventors: Alexander Kjeldaas, Serge Lachapelle
-
Patent number: 8630208Abstract: In general, this disclosure describes techniques for performing two-way muting of participant communication devices in a real-time communication session. An example method includes establishing, by a host device, a real-time communication session between a plurality of communication devices, detecting a muting condition between a first communication device and a second communication device of the plurality, and responsive to detecting the muting condition, muting one or more first audio signals that are output by the first communication device where the host device refrains from sending the one or more first audio signals to the second communication device. The method also includes, responsive to detecting the muting condition, muting one or more second audio signals that are output by the second communication device where the host device refrains from sending the one or more second audio signals to the first communication device.Type: GrantFiled: November 6, 2012Date of Patent: January 14, 2014Assignee: Google Inc.Inventor: Alexander Kjeldaas
-
Patent number: 8593501Abstract: In general, this disclosure describes techniques for providing identification, such as a name, to participants of a communication session. In one example, a method includes establishing, by a communication device, a communication session between a plurality of devices, wherein a first device of the plurality of devices is identified during the communication session by an identifier associated with a network connection of the first device to the communication session. The method also includes detecting a spoken name associated with the first device that is connected to the communication session and transforming the detected spoken name into at least one character string. The method further includes sending, by the communication device, a representation of the at least one character string to at least a second device of the plurality of devices for display by the second device.Type: GrantFiled: October 24, 2012Date of Patent: November 26, 2013Assignee: Google Inc.Inventor: Alexander Kjeldaas
-
Patent number: 8392180Abstract: In general, the techniques are described for adjusting audio gain levels for multi-talker audio. In one example, an audio system monitors an audio stream for the presence of a new talker. Upon identifying a new talker, the system determines whether the new talker is a first-time talker. For a first-time talker, the system executes a fast-attack/decay automatic gain control (AGC) algorithm to quickly determine a gain value for the first-time talker. The system additionally executes standard AGC techniques to refine the gain for the first-time talker while the first-time talker continues speaking. When a steady state within a decibel threshold is attained using standard AGC for the first-time talker, the system stores the steady state gain for the first-time talker to storage. Upon identifying a previously-identified talker, the system retrieves from storage the steady state gain for the talker and applies the steady state gain to the audio stream.Type: GrantFiled: May 18, 2012Date of Patent: March 5, 2013Assignee: Google Inc.Inventors: Serge Lachapelle, Alexander Kjeldaas
-
Patent number: 8185387Abstract: In general, the techniques are described for adjusting audio gain levels for multi-talker audio. In one example, an audio system monitors an audio stream for the presence of a new talker. Upon identifying a new talker, the system determines whether the new talker is a first-time talker. For a first-time talker, the system executes a fast-attack/decay automatic gain control (AGC) algorithm to quickly determine a gain value for the first-time talker. The system additionally executes standard AGC techniques to refine the gain for the first-time talker while the first-time talker continues speaking. When a steady state within a decibel threshold is attained using standard AGC for the first-time talker, the system stores the steady state gain for the first-time talker to storage. Upon identifying a previously-identified talker, the system retrieves from storage the steady state gain for the talker and applies the steady state gain to the audio stream.Type: GrantFiled: November 14, 2011Date of Patent: May 22, 2012Assignee: Google Inc.Inventors: Serge Lachapelle, Alexander Kjeldaas