User Tools

Site Tools


eg-speech

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
Next revision Both sides next revision
eg-speech [2018/08/08 20:49]
waltminer [Table]
eg-speech [2018/11/28 15:06]
waltminer
Line 11: Line 11:
   * Goal is to have a straw man by the AMM in Tokyo, Feb. 20-21 with work completed by ALS in June.   * Goal is to have a straw man by the AMM in Tokyo, Feb. 20-21 with work completed by ALS in June.
   * Next step is to set up a call with the engineers from all three companies who will be working on this.   * Next step is to set up a call with the engineers from all three companies who will be working on this.
-  * + 
 +===== Architecture and Design Documents ==== 
 + 
 +[[eg-speech:​architecture|Architecture]]
  
 ===== Meetings ===== ===== Meetings =====
-Meetings of the Speech EG are held every other Wednesday. Meeting time is 15:00 UTC. The upcoming schedule can be found below. ​+Meetings of the Speech EG are held every other Wednesday. Meeting time is 14:00 UTC. The upcoming schedule can be found below. ​
  
 Please join my meeting from your computer, tablet or smartphone.\\ ​ Please join my meeting from your computer, tablet or smartphone.\\ ​
Line 48: Line 51:
  
 ---------- ----------
 +
 +==== November 31, 2018 ====
 +Attendees: Walt, Jan-Simon, Paul, Naveen, Tanikawa, Shotaro, Kusakabe, Kurokawa, Fulup, Christian, Imamura, Supriya, Arijit
 +
 +   * Documents now in Confluence ​
 +     * Architecture : https://​confluence.automotivelinux.org/​display/​SpeechArch/​Speech+EG+Architecture
 +     * CES Project ​ : https://​confluence.automotivelinux.org/​display/​SpeechArch/​Speech+EG%27s+CES+2019+Project
 +
 +
 +
 +
 +==== October 31, 2018 ====
 +Attendees: Walt, Jan-Simon, Imamura, Michael F, Mike C, Ricardo, Arijit, Christian B, Thierry, Paul, Christian G, Naveen
 +
 +   * Documents now in Confluence ​
 +     * Architecture : https://​confluence.automotivelinux.org/​display/​SpeechArch/​Speech+EG+Architecture
 +     * CES Project ​ : https://​confluence.automotivelinux.org/​display/​SpeechArch/​Speech+EG%27s+CES+2019+Project
 +
 +   * Nuance announced that they need to pull people off of the CES demo work to support customer projects so they will not participate in the CES demo.  ​
 +   * Reviewed Amazon gerrit submission ​ https://​gerrit.automotivelinux.org/​gerrit/#/​c/​17877/ ​
 +
 +
 +
 +==== October 31, 2018 ====
 +Attendees: Walt, Jan-Simon, Fulup, Paul, Adam, Imamura, Arijit, Naveen, Michael F, Mike C, Ricardo, Lily the barking pug, 
 +
 +   * Documents now in Confluence ​
 +     * Architecture : https://​confluence.automotivelinux.org/​display/​SpeechArch/​Speech+EG+Architecture
 +     * CES Project ​ : https://​confluence.automotivelinux.org/​display/​SpeechArch/​Speech+EG%27s+CES+2019+Project
 +
 +   * Prototype microphones from Microchip were distributed to IoT.bzh, Amazon, and Nuance in Dresden
 +
 +   * Questions from Naveen'​s email 
 +     - CES project 2019 overall plan. What will be the setup of the green boxes and how will the box running speech framework fit into the overall demo?
 +     - Will there be a car mockup? Is it possible to do car control capabilities?​ Like climate control or locking doors.
 +       - Standard green machine set up with fans and HVAC actuators.
 +     - What is the process for submitting code to AGL?  Is there a branch that we should submit code for review to or is it master?
 +     - Should we host the High level voice service code in a public github and have Amazon, Nuance and IOT.BZH as committees? And submit a recipe to AGL Gerrit?
 +     - Once  the high level voice service code is in github we will not be immediately ready to open source Alexa voice agent code. So we will have to provide binary to Fulup'​s team for working on app integration. Will that work?
 +     - Fulup, do you have the tool chain for green box to compile the high level voice service code? What is the hardware spec of the green box?
 +     - Fulup, which apps are you planning to integrate immediately for CES? Will it include Navigation app?
 +
 +
 +==== October 17, 2018 ====
 +Attendees: ​ Canceled due to AMM in Dresden. ​
 +
 +==== October 3, 2018 ====
 +Attendees: ​
 +
 +LF: Walt, <​del>​Jan-Simon</​del>​ \\
 +Nuance: <​del>​Christian</​del>,​ Paul Purcell, <​del>​Andrew</​del>,​ <​del>​Adam</​del>,​ Mike C, <​del>​Vince</​del>,​ Arijit, Matthew Tundo\\gggggf
 +Amazon: <​del>​Premal,​ Ankur,</​del>​ Naveen, <​del>​Kamal,​ Alain</​del>​\\
 +NTT Data MSE: Imamura \\
 +Denso Ten: <​del>​Kusakabe</​del>​\\
 +Microchip: <​del>​Michael</​del>,​ <​del>​Christian</​del>​ \\
 +IoT.bzh: Stephane, Fulup, Sebastien \\
 +Konsulko: <​del>​Matt Porter, Scott, M,</​del>​ <​del>​Matt Ranostay</​del>​.\\
 +
 +   * Documents now in Confluence ​
 +     * Architecture : https://​confluence.automotivelinux.org/​display/​SpeechArch/​Speech+EG+Architecture
 +     * CES Project ​ : https://​confluence.automotivelinux.org/​display/​SpeechArch/​Speech+EG%27s+CES+2019+Project
 +
 +   * Reviewed questions and comments about latest Amazon proposal (v1.3). ​
 +     * Wake word engine when multiple agents are present
 +     * PTT versus wake word versus a mixed mode
 +     * Multi-modal interactions when engaged in a voice session
 +
 +   * Next Steps (for F2F next week)
 +     * Further review of high level architecture contained in the document ​
 +     * Review the proposed API in the document
 +     * Start to define support binding APIs both reuse of existing ones and new ones that may be required
 +     * Input audio architecture ​
 +
 +
 +
 +==== September 19, 2018 ====
 +Attendees:  ​
 +
 +LF: Walt, <​del>​Jan-Simon</​del>​ \\
 +Nuance: Christian, Paul Purcell, <​del>​Andrew</​del>,​ Adam, Mike C, <​del>​Vince</​del>,​ Arijit, Matthew Tundo\\
 +Amazon: Premal, Ankur, Naveen, Kamal, Alain\\
 +NTT Data MSE: Imamura \\
 +Denso Ten: <​del>​Kusakabe</​del>​\\
 +Microchip: <​del>​Michael</​del>,​ <​del>​Christian</​del>​ \\
 +IoT.bzh: Stephane, Fulup \\
 +Konsulko: Matt Porter, Scott, M, <​del>​Matt Ranostay</​del>​.\\
 +
 +   * Walt working on getting a Confluence site set up for AGL 
 +   * F2F meeting outcome from Santa Clara should be sent out publicly sometime this week. 
 +
 +   * Reviewed questions and comments about latest Amazon proposal (v1.1). ​
 +     * Wake word engine when multiple agents are present
 +     * PTT versus wake word versus a mixed mode
 +     * Multi-modal interactions when engaged in a voice session
 +
 +   * Next Steps (for F2F next week)
 +     * Further review of high level architecture contained in the document ​
 +     * Review the proposed API in the document
 +     * Start to define support binding APIs both reuse of existing ones and new ones that may be required
 +     * Input audio architecture ​
 +
 +
 +
 +==== September 6, 2018 ====
 +Attendees:  ​
 +LF: Walt, <​del>​Jan-Simon</​del>​ \\
 +Nuance: Christian, Paul Purcell, Andrew, Adam, Mike C, Vince, Arijit, Matthew Tundo\\
 +Amazon: Premal, Ankur, Naveen, Kamal, Alain\\
 +NTT Data MSE: Imamura \\
 +Denso Ten: Kusakabe\\
 +Microchip: Michael, Christian \\
 +IoT.bzh: Stephane, Fulup \\
 +Konsulko: Matt Porter, Scott, M, <​del>​Matt Ranostay</​del>​.\\
 +
 +   * Reviewed questions and comments about latest Amazon proposal (v1.1). ​
 +     * Wake word engine when multiple agents are present
 +     * PTT versus wake word versus a mixed mode
 +     * Multi-modal interactions when engaged in a voice session
 +
 +   * Next Steps (for F2F next week)
 +     * Further review of high level architecture contained in the document ​
 +     * Review the proposed API in the document
 +     * Start to define support binding APIs both reuse of existing ones and new ones that may be required
 +     * Input audio architecture ​
 +
 +
  
 ==== September 5, 2018 ==== ==== September 5, 2018 ====
 Attendees: ​ //Upcoming Meeting// ​ Attendees: ​ //Upcoming Meeting// ​
 +
 +LF: Walt, <​del>​Jan-Simon</​del>​ \\
 +Nuance: Christian, Paul Purcell, Andrew, Adam, Mike C, Vince, Arijit, Matthew Tundo\\
 +<​del>​Amazon:​ Premal, Ankur, Naveen, Kamal, Alain</​del>​\\
 +NTT Data MSE: Imamura \\
 +Denso Ten: Kusakabe\\
 +Microchip: Michael, Christian \\
 +IoT.bzh: Stephane, Fulup \\
 +Konsulko: Matt Porter, Scott, M, <​del>​Matt Ranostay</​del>​.\\
 +
 +
 +Notes:
 +
 +Nuance still discussing internally about releasing their API. Christian working with AGL App FW and working with writing an AGL Service layer in github (https://​github.com/​Nuance-Mobility/​agl-speech-interface)
 +   * Not wired up to a speech or TTS engine, more of a loop back test. 
 +   * Will send the link to Konsulko and IoT.bzh to review the API for suggestions. ​
 +     * Done.
 +   * List of AGL services available can be seen at https://​git.automotivelinux.org/​
 +   * Need to figure out consent and privacy issues with AGL Identity Agent.
 +     * No update.
 +   * How to manage grammar and natural language APIs and split between services and apps? 
 +   * How to integrate cloud speech applications?​
 +     * Example: "Find me the closest pizza place" is processed in the cloud and the location and name are returned to the ECU.  How is this then transmitted to the POI and/or navi app?
 +   * Sample config from softmixer https://​github.com/​iotbzh/​4a-softmixer/​blob/​master/​conf.d/​project/​lua.d/​smixer-test-simple.lua
 +     * Update 7/25: update for FF use 8-channel CSL usb dac, about to land in gerrit.
 +   * Starting a demo project internally led by Paul. End of Aug early Sep they plan to have a design document together internally and will be ready with any questions/​issue. Would be a good idea to target the [[agl-distro:​sep2018-f2f|Sep F2F in Santa Clara]] to resolve issues with the design. ​
 +   * Information on IRC, mail list etc is available at [[start:​getting-started|Getting Started with AGL]]
 +   * Supported hardware can be found at [[agl-distro#​supported_hardware|AGL Distribution]]
 +   * 8/8 
 +     * Arijit received the M3 hardware and was able to get it running. Building a "Hello World" sample application using Virtual Box and M3 hardware. ​
 +   * 8/21
 +     * Nuance email list Automotive-Grade-Linux@nuance.com
 +   * 9/5
 +     * Matt Tundo working on test audio application. Having trouble getting microphone capture via ALSA and 4a. 
 +     * Document for getting AGL working in native Linux http://​docs.automotivelinux.org/​docs/​devguides/​en/​dev/​reference/​host-configuration/​docs/​1_Prerequisites.html
 +
 +
 +Amazon looking at releasing a possible API in June and starting to work with the AGL App FW. 
 +   * Starting to look at AGL App FW binder implementation using audio HAL as a reference. Will work with IoT.bzh on how to put the configuration together. ​
 +     * Update 7/25: reviewing above draft, will share ideas/​design to run multiple engines in parallel. No timeline, yet. Will review internally and present in next call.
 +   * Would like to put together an architecture picture based on the white board drawings from February AMM to see how the API fits into AGL overall.  ​
 +   * 8/8 
 +     * Naveen presented some use cases and an architecture diagram that Amazon has been working on internally. Received good feedback from the team. Naveen and his team will update their internal wiki and present again at the next meeting. Will look into getting the info onto the AGL wiki after that. 
 +
 +Microchip - AGL USB microphone front-end. Michael is working with MicroSemi on getting hardware available that is already available for Amazon Alexa. Would like to have a prototype available for the AMM in Dresden. Microchip plans to provide the HAL for the microphone. ​
 +    * Update 7/25:
 +      * Received hardware from MicroSemi, Alexa stack already working.
 +      * Hardware is mic+dsp connected to rpi running the stack
 +      * Plan: frontend should be connected over USB, integrated with 4a (hal) and interacting with the stack
 +        * Stack needs to pick-up conditioned signal (near/​far/​noise-cancelling) through alsa device
 +        * Michrochip will provide the hal for 4a
 +        * Michael: Interest to extend the API for beamforming,​ multiple seats, "1 channel per seat" ?
 +      * Update 8/22 
 +        * Received five eval kits from MicroSemi. So far so good. Prototypes will be delivered for AMM. 
 +      * Update 9/5 
 +        * Dresden time-frame will some units ready for evaluation. After that they will mass-produce a since PCB devices that will be readily available for purchase. ​
 +
 +Action item:
 +   * Walt to set up extra call for later this week. 
 +
  
 ==== August 22, 2018 ==== ==== August 22, 2018 ====
 Attendees: //Upcoming Meeting// ​ Attendees: //Upcoming Meeting// ​
 +
 +LF: Walt, Jan-Simon \\
 +Nuance: Christian</​del>,​ Paul Purcell, <​del>​Mike C.</​del>,​ <​del>​Vince</​del>,​ Arijit, Matthew\\
 +Amazon: <​del>​Premal</​del>,​ Ankur, <​del>​Naveen</​del>,​ Kamal, Alain\\
 +NTT Data MSE: Imamura \\
 +Denso Ten: Kusakabe\\
 +Microchip: Michael, <​del>​Christian</​del>​ \\
 +IoT.bzh: <​del>​Stephane</​del>,​ <​del>​Fulup</​del>​ \\
 +Konsulko: <​del>​Matt P.</​del>,​ <​del>​Matt R.</​del>​\\
 +Myscript: <​del>​Olivier,​ Etienne</​del>​ \\
 +
 +
 +Notes:
 +
 +Nuance still discussing internally about releasing their API. Christian working with AGL App FW and working with writing an AGL Service layer in github (https://​github.com/​Nuance-Mobility/​agl-speech-interface)
 +   * Not wired up to a speech or TTS engine, more of a loop back test. 
 +   * Will send the link to Konsulko and IoT.bzh to review the API for suggestions. ​
 +     * Done.
 +   * List of AGL services available can be seen at https://​git.automotivelinux.org/​
 +   * Need to figure out consent and privacy issues with AGL Identity Agent.
 +     * No update.
 +   * How to manage grammar and natural language APIs and split between services and apps? 
 +   * How to integrate cloud speech applications?​
 +     * Example: "Find me the closest pizza place" is processed in the cloud and the location and name are returned to the ECU.  How is this then transmitted to the POI and/or navi app?
 +   * Sample config from softmixer https://​github.com/​iotbzh/​4a-softmixer/​blob/​master/​conf.d/​project/​lua.d/​smixer-test-simple.lua
 +     * Update 7/25: update for FF use 8-channel CSL usb dac, about to land in gerrit.
 +   * Starting a demo project internally led by Paul. End of Aug early Sep they plan to have a design document together internally and will be ready with any questions/​issue. Would be a good idea to target the [[agl-distro:​sep2018-f2f|Sep F2F in Santa Clara]] to resolve issues with the design. ​
 +   * Information on IRC, mail list etc is available at [[start:​getting-started|Getting Started with AGL]]
 +   * Supported hardware can be found at [[agl-distro#​supported_hardware|AGL Distribution]]
 +   * 8/8 
 +     * Arijit received the M3 hardware and was able to get it running. Building a "Hello World" sample application using Virtual Box and M3 hardware. ​
 +   * 8/21
 +     * Nuance email list Automotive-Grade-Linux@nuance.com
 +     ​* ​
 +
 +
 +Amazon looking at releasing a possible API in June and starting to work with the AGL App FW. 
 +   * Starting to look at AGL App FW binder implementation using audio HAL as a reference. Will work with IoT.bzh on how to put the configuration together. ​
 +     * Update 7/25: reviewing above draft, will share ideas/​design to run multiple engines in parallel. No timeline, yet. Will review internally and present in next call.
 +   * Would like to put together an architecture picture based on the white board drawings from February AMM to see how the API fits into AGL overall.  ​
 +   * 8/8 
 +     * Naveen presented some use cases and an architecture diagram that Amazon has been working on internally. Received good feedback from the team. Naveen and his team will update their internal wiki and present again at the next meeting. Will look into getting the info onto the AGL wiki after that. 
 +
 +Microchip - AGL USB microphone front-end. Michael is working with MicroSemi on getting hardware available that is already available for Amazon Alexa. Would like to have a prototype available for the AMM in Dresden. Microchip plans to provide the HAL for the microphone. ​
 +    * Update 7/25:
 +      * Received hardware from MicroSemi, Alexa stack already working.
 +      * Hardware is mic+dsp connected to rpi running the stack
 +      * Plan: frontend should be connected over USB, integrated with 4a (hal) and interacting with the stack
 +        * Stack needs to pick-up conditioned signal (near/​far/​noise-cancelling) through alsa device
 +        * Michrochip will provide the hal for 4a
 +        * Michael: Interest to extend the API for beamforming,​ multiple seats, "1 channel per seat" ?
 +      * Update 8/22 
 +        * Received five eval kits from MicroSemi. So far so good. Prototypes will be delivered for AMM. 
 +
 +     * Question from Nuance about audio streaming:
 +       * esoundlib - do you need special calls to stream audio
 +       * Fulup: no, reply of 4a role request is the alsa device to write to
 +       * 4a-play /​usr/​share/​4a/​media/​Happy_MBB_75.ogg (only script)
 +
 +Action item:
 +   ​* ​
  
  
eg-speech.txt · Last modified: 2021/11/23 15:57 by wminer