HoundHouse

This is a C++ client that interfaces with the Houndify platform and is meant to be run on a Raspberry Pi 3 B+ running Ubuntu 18.04 LTS.

I personally work on this repository via Elementary OS 5.0 (which runs Ubuntu 18.04 LTS under the hood) on a Dell XPS 13 9370. In addition I'm also using a handful of my personal Raspberry Pi 3 B+s that are running Ubuntu 18.04 LTS. To create Ubuntu 18.04 SD cards for use with my Pi 3 B+s I followed the instructions in the second post in this thread on the Raspberry Pi forums.

Hardware Recommendations

These are the prodcuts that I recommend and they are also the ones that I will test. This must be kept in mind when developing any and all features for this repository; however note that the Simple Drivers in this repository should be able to function with any ALSA compatible microphone and speaker.

Name	Function	Link To Source	Purchase Link
Raspberry Pi 3 B+	THE HEART OF IT ALL!!	Source	Purchase
ReSpeaker 4-Mic Array	Microphone	Source	Purchase
Pibow 3 B+ Coupé Flotilla	Case/Protection	Source	Purchase
Raspberry Pi Universal Power Supply	Power Supply	Source	Purchase
Sony XB10	Speaker	Source	Purchase
6 Inch 3.5mm audio cable	Pi to Speaker link		Purchase

Note: Yes, initially this project uses a speaker via the audio jack; however it may not in the future. The only stipulation about what speaker can be used is that it must be ALSA compatible. This could change in the future.

Getting started

These are the steps you'll need to take to get up and running with this repository!

1. Clone this repository somewhere nice to work with

    git clone https://github.com/TylerSeanRau/HoundHouse.git ~/Downloads/HoundHouse
   

   • I use ~/Downloads/ throughout this README but feel free to use any directory you'd like, just be sure you always use the same substitute directory in its place.

2. Download and build the Houndify C++ Standard SDK

   1. Download the C++ Standard SDK from the Houndify C++ SDKs page

   2. Extract the file you downloaded in step one.

       tar -xvf HoundCpp.tar.gz -C ~/Downloads
      

   3. If you're working with this repository on an actual Raspberry Pi 3 B+ then you'll need to do a couple extra steps before making or you'll run into memory issues, both during compiling and during linking.

      • I've been running the following commands:

          cd ~/Downloads/HoundCpp
          for i in CFLAGS= CPP_FLAGS=; do for j in `grep -R "^$i" -l`; do sed -i "s/^$i/$i -Os -Wl,--no-keep-memory -Wl,--reduce-memory-overhead /g" $j; done; done
          for i in `grep -R -F '$(LINK_CPP) -o $(LINK_FLAGS)' -l`; do sed -i 's/^LINK_FLAGS=/LINK_FLAGS= -Os -Wl,--no-keep-memory -Wl,--reduce-memory-overhead /g; s/$(LINK_CPP) -o $(LINK_FLAGS)/$(LINK_CPP) $(LINK_FLAGS) -o/g' $i; done
        

   4. Run make

      • If you're on Linux with a decent machine running an 8 thread CPU

          cd ~/Downloads/HoundCpp
          make -j 8
        

      • If you're on OSX with a decent machine running an 8 thread CPU

          cd ~/Downloads/HoundCpp
          make -j 8 -f Makefile.OSX
        

      • If you're on a Raspberry Pi 3 B+ and running bash

          cd ~/Downloads/HoundCpp
          nohup bash -c 'make -j 2' > ../makelog 2>&1 & disown
        

3. Obtain the "Ok Hound" phrase spotter

   • A. If you're running Linux x86-64 follow these instructions

     1. First create a directory for these files in your local clone of this repository.

         mkdir -p ~/Downloads/HoundHouse/OkHound/linux-x86_64
        

     2. Download the "Ok Hound" Phrase Spotter for Node.js Linux x86-64 (1.0.0) from the Houndify Javascript SDKs page

        • DO NOT DOWNLOAD TO THE DIRECTORY YOU CREATED IN 3.A.1

     3. Extract the file you just downloaded

         tar -xvf okhound-nodejs-1.0.0-linux-x86_64.tar.gz -C ~/Downloads
        

     4. Extract the okhound library

         tar -xvf ~/Downloads/okhound-linux-x86_64/okhound-1.0.0.tgz -C ~/Downloads/okhound-linux-x86_64/
        

     5. Copy the needed files to the folder you created in step 3.A.1

         cp ~/Downloads/okhound-linux-x86_64/package/libPhraseSpotter.a ~/Downloads/okhound-linux-x86_64/package/PhraseSpotterAPI.h ~/Downloads/okhound-linux-x86_64/package/LICENSE* ~/Downloads/HoundHouse/OkHound/linux-x86_64
        

     6. Clean up files that are no longer needed

         rm -rf ~/Downloads/okhound-linux-x86_64/ ~/Downloads/okhound-nodejs-1.0.0-linux-x86_64.tar.gz
        

   • B. If you're running on a Raspberry Pi 3 B+ follow these instructions

     1. First create a directory for these files in your local clone of this repository.

         mkdir -p ~/Downloads/HoundHouse/OkHound/raspberrypi
        

     2. Download the "Ok Hound" Phrase Spotter for Node.js Raspberry Pi 2 & 3 (1.0.0) from the Houndify Javascript SDKs page

        • DO NOT DOWNLOAD TO THE DIRECTORY YOU CREATED IN 3.B.1

     3. Extract the file you just downloaded

         tar -xvf okhound-nodejs-1.0.0-pi2+3.tar.gz -C ~/Downloads
        

     4. Extract the okhound library

         tar -xvf ~/Downloads/okhound-pi2\&3/okhound-1.0.0.tgz -C ~/Downloads/okhound-pi2\&3/
        

     5. Copy the needed files to the folder you created in step 3.B.1

         cp ~/Downloads/okhound-pi2\&3/package/libPhraseSpotter.a ~/Downloads/okhound-pi2\&3/package/PhraseSpotterAPI.h ~/Downloads/okhound-pi2\&3/package/LICENSE* ~/Downloads/HoundHouse/OkHound/raspberrypi
        

     6. Clean up files that are no longer needed

         rm -rf ~/Downloads/okhound-pi2\&3/ ~/Downloads/okhound-nodejs-1.0.0-pi2+3.tar.gz
        

4. Extract useful utilities from HoundifyExplorer

   • Note, at some point I may create these utilities myself but for now the easiest thing to do was to use the files bundled with HoundifyExplorer

   1. First create a directory for these files in your local clone of this repository.

       mkdir -p ~/Downloads/HoundHouse/HoundifyExplorerUtils
      

   2. Download the Houndify Explorer Source from the Houndify Houndify Explorer page

   3. Extract the file you just downloaded.

       tar -xvf HoundifyExplorer.tar.gz -C ~/Downloads
      

   4. Copy utilities used by this repo

       cp ~/Downloads/HoundifyExplorer/HoundClientALSAAudio.cpp ~/Downloads/HoundHouse/HoundifyExplorerUtils
       cp ~/Downloads/HoundifyExplorer/HoundClientALSAAudio.h ~/Downloads/HoundHouse/HoundifyExplorerUtils
       cp ~/Downloads/HoundifyExplorer/PlatformSpecificClientCapabilities.h ~/Downloads/HoundHouse/HoundifyExplorerUtils
       cp ~/Downloads/HoundifyExplorer/PlatformSpecificClientCapabilitiesLinux.cpp ~/Downloads/HoundHouse/HoundifyExplorerUtils
       cp ~/Downloads/HoundifyExplorer/HoundAudioVerifier.h ~/Downloads/HoundHouse/HoundifyExplorerUtils
      

   5. OPTIONAL Clean up files that are no longer needed

       rm -rf ~/Downloads/HoundifyExplorer ~/Downloads/HoundifyExplorer.tar.gz
      

      • I strongly recommend you keep HoundifyExplorer and read its README. It's an invaluable debugging tool.

Building the SmartDevice libray

If you would like to take advantage of the smart devices libray that is included here then you'll need to build the library.

To do so cd to SmartDevices/lib and run make this will build all the smart devices libraries and unify them into a single SmartDevices library located SmartDevices/lib/libSmartDevices.a. It should be straight forward to change makefiles to only use the specific subsection you'd like included in your project. (If space is a concern.)

The simple demos make files should allow -DWITHALLSMARTDEVICESUPPORT to be added to DEFINE_FEATURE_FLAGS and then the appopriate -I/-L/-l flags will be added to the different steps. Note, the -I flag goes from the the SmartDevices/include directory. So you'll need to write #include "*.h" paths with the appopriate path. (Or add the desired include folder to the -I.)

Simple Demos

SimplePhraseSpotter

SimplePhraseSpotter.cpp is a basic driver that uses an ALSA compatible microphone to watch for utterances of "Ok Hound"

SimpleAudioQueriesOnlyWrittenResponse

SimpleAudioQueriesOnlyWrittenResponse.cpp is another basic driver that uses an ALSA compatible microphone to listen for utterences of "Ok Hound" and responds only with text on the screen. * Note, it would not take much modification to make this driver support audio responses, most of the required changes would be inside the classes it uses not in the main .cpp file