14 Replies Latest reply on Feb 15, 2019 2:31 PM by Jan Cumps

    How Would You RoadTest the Matrix Voice


      Voice applications are gaining popularity. The apps can range from far field voice captures, acoustic source localization, beamforming, noise suppression, acoustic echo cancellation, voice recognition integration, and more.


      So, it should be no surprise to the RoadTester group that I would be approached to roadtest a voice-app dev board. Enter, the Matrix Voice.


      Last year, we roadtested the Matrix Creator for the Raspberry Pi , which is an expansion daughter board for the Raspberry Pi. The Matrix Voice has the same circular shape as the Creator, but is a simple, easy-to-use, open-source voice recognition platform. It's 3.14 inches in diameter with a radial array of 7 MEMS microphones connected to a Xilinx Spartan 6 FPGA and 64 Mbit SDRAM with 18 RGBW LED's and 16 GPIO pins.


      There are two versions of the Matrix Voice: one version works with the Raspberry Pi, and the other version has an onboard ESP32 Wi-Fi/BT module with a 32 bit microcontroller, so it can work as a standalone device. Since I know my Roadtesters love the tech details, here's a few specs:


      • FPGA : Xilinx Spartan 6 FPGA XC6SLX9
      • 18 RGBW LEDs (link)
      • 8 MEMS audio sensor MP34DB02
      • 16 external GPIO. These are connected to FPGA so they can implement any digital
      • interface e.eg PWM, Servo, UART, I2C etc.
      • Serial Flash 64MBIT - MX25L6406E
      • DDR2 SDRAM 512MBIT - MT47H32M16
      • 3W Stereo Class-D Audio Amplifier and Class-AB Headphone Driver - PAM8019
      • 3.5mm audio output jack


      Here's a few tech resources for you to examine:


      So here's my question: How Would You roadTest the Matrix Voice? What would you build with the Matrix Voice? What would you like to know about the Matrix Voice. Leave your thoughts and comments below. Thanks.


      Randall Scasny

      RoadTest Program Manager



      Before I go.....


      I saw this video on youtube where the Matrix Voice was used with Alexa. Some asked:


      "Alexa, how much wood can a wood chuck chuck if a wood chuck could chuck wood? Watch the video to find out:


        • Re: How Would You RoadTest the Matrix Voice
          Vimarsh ‍

          Matrix voice is a good add on to the RPi. There are many such hats but this one is the best of it due to that FPGA.


          The advantage is we can make a home automation system, a google home for ourselves or make our own assistants with things like snips.ai


          It is also smaller than the creator. It also does not have esp32 moreover the sensors are lacking in the simple version. The one with esp32 can be more helpful as it can be programmed standalone and so we can also add functionalities which are difficult with FPGA or simply not possible. Even the creator does not have ESP32 a very common chip to program for standalone operation.


          But being small, and the awesome projects and manufacturer support we can create cool home automation projects.

          HA is quite common and also has many exixting solutions so we need to be creative to think where else it can be used? Probably in some alarming system or something else.


          We can test it by creating an all-in-one project as that is what the board is capable to do and we can also experiment with FPGA which have good documentation.




          2 of 2 people found this helpful
          • Re: How Would You RoadTest the Matrix Voice
            Jan Cumps

            For those that want to re-program the on-board FPGA (not needed to work with the Creator but it can be done), a heads-up:

            The Xilinx ISE IDE that you use to work with the Spartan 6 is not compatible with Windows 10.  You need to have access to an older Windows version or use the virtual machine with ISE pre-installed that Xilinx released.

            • Re: How Would You RoadTest the Matrix Voice

              Hi Randall,


              When I saw this device I had two thoughts.

              One was to use it for direction finding for thunder to alert someone about the direction of an approaching thunder storm.

              The other thought was for tracking individuals in a meeting room. I am assuming that you could have everyone identify themselves and use the microphones to map each person. You could then copy statements made by each person during the meeting using the previous location information to identify the talker without needing to do voice comparison. You could also use the voice comparison to track the movement of a specific speaker within the room.



              • Re: How Would You RoadTest the Matrix Voice

                It has been a long while since I last tinkered with psychoacoustic masking and audio steganography, but reading about the Matrix Voice awoke some of those ideas.


                I'm thinking the Matrix Voice could be adapted to perform real time audio steganographic decoding. Not that the development effort to achieve real time steganographic decoding would be trivial, but it seems at first blush to have the necessary computational horsepower.


                On another front: How about a near real time translator that converts spoken word in one language into written text in another language?  Maybe most of the heavy lifting could be done by Google services with the Matrix Voice handling digitization, formatting and upload of voice data.  Just a thought.


                Both of these ideas go well beyond the scope of a Road Test, but would make fun and challenging multi-month projects.

                • Re: How Would You RoadTest the Matrix Voice

                  Matrix voice is a very good add on board for projects using voice interfrace with raspberry pi . It contains good microphone and we can use it to make project something for home automation and try the different sensors inbuilt in it .The standalone version would be helpful where there is need of low cost project.


                  I would like to road test it by making a personal voice assistant for us using raspberry pi and matrix voice and use the data of inbuilt sensors in it  and try to put my hands on fpga.



                  Aabhas Senapati

                  2 of 2 people found this helpful
                  • Re: How Would You RoadTest the Matrix Voice

                    It would be interesting to add some directional baffles and use the device for determining sound source direction.  With haptic feedback, it could be a good aid for the deaf.


                    • Re: How Would You RoadTest the Matrix Voice

                      I don't have any esoteric needs, but I have 2 mundane applications that have been bugging me for a long time:

                      1. A long range microphone I can locate at my large screen TV and use for Skype calls while sitting across the room. (A wired mic running across the living room just doesn't cut it.)
                      2. A mic for my camera that will equalize my voice volume as I blog - no matter how far away I move (up to 8 feet). Wearing a wired mic often gets in the way while doing a demo. Even a wireless Lavalier is cumbersome and distracting in a video.

                      My attempts to solve these applications so far have been dismal. (I am open to suggestions)

                      1 of 1 people found this helpful
                        • Re: How Would You RoadTest the Matrix Voice

                          Perhaps overhead choir mics above your work area. If you feed them through a compressor you can even off the levels as you move around - set the dynamics to increase compression as you get closer and back the compression off as you move further away.


                          A number of conference table type applications are moving to 'beam forming' ceiling microphone arrays. However ceilings often have a lot of noise with ventilation and projector fans.

                          2 of 2 people found this helpful