1. Integrating and using speech recognition and transcription
In this tutorial series, you will create a Mixed Reality application that explores the use of Azure Speech Services with the HoloLens 2. When you complete this tutorial series, you will be able to use your device's microphone to transcribe speech to text in real time, translate your speech into other languages, and leverage the Intent recognition feature to understand voice commands using artificial intelligence.
- Learn how to integrate Azure Speech Services with a HoloLens 2 application
- Learn how to use speech recognition to transcribe text
If you have not completed the Getting started tutorials series yet, it's recommended that you complete those tutorials first.
- A Windows 10 PC configured with the correct tools installed
- Windows 10 SDK 10.0.18362.0 or later
- Some basic C# programming ability
- A HoloLens 2 device configured for development
- Unity Hub with Unity 2019.2.X installed and the Universal Windows Platform Build Support module added
The recommended Unity version for this tutorial series is Unity 2019.2.X. This supersedes any Unity version requirements or recommendations stated in the prerequisites linked above.
Creating and preparing the Unity project
In this section, you will create a new Unity project and get it ready for MRTK development.
- Creating the Unity project and give it a suitable name, for example, MRTK Tutorials
- Switching the build platform
- Importing the TextMeshPro Essential Resources
- Importing the Mixed Reality Toolkit
- Configuring the Unity project
- Creating and configuring the scene and give the scene a suitable name, for example, AzureSpeechServices
Then follow the Changing the Spatial Awareness Display Option instructions to change the MRTK configuration profile for your scene to the DefaultHoloLens2ConfigurationProfile and change the display options for the spatial awareness mesh to Occlusion.
Configuring the speech commands start behavior
Because you will use the Speech SDK for speech recognition and transcription you need to configure the MRTK Speech Commands so they do not interfere with the Speech SDK functionality. To achieve this you can change the speech commands start behavior from Auto Start to Manual Start.
With the MixedRealityToolkit object selected in the Hierarchy window, in the Inspector window, select the Input tab, clone the DefaultHoloLens2InputSystemProfile and the DefaultMixedRealitySpeechCommandsProfile, and then change the speech commands Start Behavior to Manual Start:
For a reminder on how to clone and configure MRTK profiles, you can refer to the Configuring the Mixed Reality Toolkit profiles instructions.
Configuring the capabilities
In the Unity menu, select Edit > Project Settings... to open the Player Settings window, then locate the Player > Publishing Settings section:
In the Publishing Settings, scroll down to the Capabilities section and double-check that the InternetClient, Microphone, and SpatialPerception capabilities, which you enabled when you created the project at the beginning of the tutorial, are enabled. Then, enable the InternetClientServer and PrivateNetworkClientServer capabilities:
Importing the tutorial assets
Download and import the following Unity custom packages in the order they are listed:
- Microsoft.CognitiveServices.Speech.N.N.N.unitypackage (latest version)
For a reminder on how to import a Unity custom package, you can refer to the Importing the Mixed Reality Toolkit instructions.
After you have imported the tutorial assets your Project window should look similar to this:
Preparing the scene
In this section, you will prepare the scene by adding the tutorial prefab and configure the Lunarcom Controller (Script) component to control your scene.
In the Project window, navigate to Assets > MRTK.Tutorials.AzureSpeechServices > Prefabs folder and drag the Lunarcom prefab into the Hierarchy window to add it to your scene:
With the Lunarcom object still selected in the Hierarchy window, in the Inspector window, use the Add Component button to add the Lunarcom Controller (Script) component to the Lunarcom object:
The Lunarcom Controller (Script) component is not part of MRTK. It was provided with this tutorial's assets.
With the Lunarcom object still selected, expand it to reveal its child objects, then drag the Terminal object into the Lunarcom Controller (Script) component's Terminal field:
With the Lunarcom object still selected, expand the Terminal object to reveal its child objects, then drag the ConnectionLight object into the Lunarcom Controller (Script) component's Connection Light field and the OutputText object into the Output Text field:
With the Lunarcom object still selected, expand the Buttons object to reveal its child objects, and then in the Inspector window, expand the Buttons list, set its Size to 3, and drag the MicButton, SatelliteButton, and RocketButton objects into the Element 0, 1, and 2 fields respectively:
Connecting the Unity project to the Azure resource
To use Azure Speech Services, you need to create an Azure resource and obtain an API key for the Speech Service. Follow the Try the Speech service for free instructions and make a note of your service region (also known as Location) and API key (also known as Key1 or Key2).
In the Hierarchy window, select the Lunarcom object, then in the Inspector window, locate the Lunarcom Controller (Script) component's Speech SDK Credentials section and configure it as follows:
- In the Speech Service API Key field, enter your API key (Key1 or Key2)
- In the Speech Service Region field, enter your service region (Location) using lowercase letters and spaces removed
Using speech recognition to transcribe speech
In the Hierarchy window, select the Lunarcom object, then in the Inspector window, use the Add Component button to add the Lunarcom Speech Recognizer (Script) component to the Lunarcom object:
The Lunarcom Speech Recognizer (Script) component is not part of MRTK. It was provided with this tutorial's assets.
If you now enter Game mode, you can test the speech recognition by first pressing the microphone button:
Then, assuming your computer has a microphone, when you say something, your speech will be transcribed on the terminal panel:
The application needs to connect to Azure, so make sure your computer/device is connected to the internet.
You have implemented speech recognition powered by Azure. Run the application on your device to ensure the feature is working properly.
In the next tutorial, you will learn how to execute commands using Azure speech recognition.