AudioCaptureHandler.h

Go to the documentation of this file.

// Copyright(c) 2024 grrimgrriefer & DZnnah, see LICENSE for details.
 
#pragma once
 
#include "CoreMinimal.h"
#include "Misc/ScopeLock.h"
#include "VoiceRunnerThread.h"
#include "VoxtaDefines.h"
 
class AudioWebSocket;
class IVoiceCapture;

class VOXTAAUDIOUTILITY_API AudioCaptureHandler
{
#pragma region public API
public:
    void RegisterSocket(TWeakPtr<AudioWebSocket> socket, int bufferMillisecondSize);

    bool TryInitializeVoiceCapture(int sampleRate = 16000, int numChannels = 1);

    void ConfigureSilenceThresholds(float micNoiseGateThreshold, float silenceDetectionThreshold, float micInputGain);

    bool TryStartVoiceCapture();

    void StopCapture();

    void ShutDown(bool alsoDestroyCaptureDevice = false);

    bool IsInputSilent() const;

    float GetDecibels() const;

    const FString& GetDeviceName() const;
 
    void SetIsTestMode(bool isTestMode);
#pragma endregion
 
#pragma region data
private:
    TArray<uint8> m_socketDataBuffer;
 
    static constexpr float DEFAULT_SILENCE_DECIBELS = -144.f;

    friend class FVoiceRunnerThread;
 
    FString m_deviceName = EMPTY_STRING;
    bool m_isCapturing = false;
    int m_bufferMillisecondSize = 0;
 
    float m_decibels = DEFAULT_SILENCE_DECIBELS;
    bool m_isTestMode = false;
    FDateTime m_lastVoiceTimestamp = FDateTime();
 
    mutable FCriticalSection m_captureGuard;
    TUniquePtr<FVoiceRunnerThread> m_voiceRunnerThread;
    TWeakPtr<AudioWebSocket> m_webSocket;
    TSharedPtr<IVoiceCapture> m_voiceCaptureDevice;
#pragma endregion
 
#pragma region private API
    void CaptureAndSendVoiceData();

    bool CaptureVoiceInternal(TArray<uint8>& voiceDataBuffer, float& decibels) const;

    void SendInternal(const TArray<uint8>& rawData) const;

    float AnalyseDecibels(const TArray<uint8>& voiceInputData, uint32 dataSize) const;

    bool IsInputSilentInternal() const;
#pragma endregion
};