MRTK V2.2 - Access Speech Command via Script

Question

In my scenario, buttons are created during runtime. These are to be clicked by a voice command. For this reason I try to find out how I can add voice commands during runtime. But I can't find any approach.

What I tried: I have extended the interface IMixedRealitySpeechSystem with two methods, RefreshRecognition and AddSpeechCommand:

/// <summary>
/// Refresh recognition after adding new commands
/// </summary>
void RefreshRecognition();

/// <summary>
/// Add command to already existing commands[]
/// </summary>
/// <param name="command"></param>
void AddSpeechCommand(SpeechCommands command);

I have implemented these in the class WindowsSpeechInputProvider: MixedRealitySpeechSystem. But there are two problems.

First: I can't get to the WindowsSpeechInputProvider. I thought I could get it by trying this:

private IMixedRealitySpeechSystem SpeechSystem
{
    get
    {
        if(_speechSystem is null)
        {
            MixedRealityServiceRegistry.TryGetService(out _speechSystem);
        }
        return _speechSystem;
    }
}

public void SomeMethod()
{
    SpeechCommands command = new SpeechCommands("TestCommand", default, default, null);
    SpeechSystem.AddSpeechCommand(command);
    SpeechSystem.RefreshRecognition();
}

But the problem is that MixedRealityServiceRegistry does not contain an instance of that service or to be precise, it is not even a service.

Second: Even if this would work, it is not a good way to go. Because with this I change the MRTK and with another upgrade to a new version, these lines are overwritten.

My Question:
So how can I access and add commands on runtime?

Answer 1

There's an open feature request to allow adding dynamic speech commands in Github: Add keywords dynamically to MRTK speech commands #6369. It is not currently possible.

This thread has some suggestions for alternative ways to approach the overall scenario. In summary, it is recommended that you use a Grammar Recognizer and use an SRGS XML file to define your speech recognition rules. Voice input in Unity and Hologram 212 has an example showing how to use it.