(Unity 5.6) (IBM Watson SDK) 将用户语音解析为文本，然后再解析为语音

Question

我正在尝试使用新的 [IBM Watson SDK for Unity] 获取用户的语音输入，将其解析为文本，然后将该文本传回 watson 的文本转语音，以便它可以重复用户输入的内容说。

我已经使用 SDK 的 Speech To Text widget[takes audio clip from microphone and parses into text], in conjunction with it's Speech display widget [将文本输入从语音转换为文本并将其显示在 canvas 场景中] 在屏幕上显示用户的语音。然后我创建了一个 UI 输入字段，将其与 SpeechDisplaywidget 的文本输出以及 texttoSpeechWidget 的输入相关联 - 然后制作了一个 UI 按钮，单击该按钮时，将文本发送到文本到语音服务。它应该播放关联的文本到语音文件，但是当我单击该按钮时，没有任何反应。 'status' 显示只是读取 "ready" 并且从不改变，整个事情没有抛出任何错误。我会粘贴代码，但我还没有编写任何代码，只是连接了所有必要的序列化字段并点击开始。我该怎么办？

抱歉含糊不清，但由于这是我的第一个问题，我无法 post 更多图片或链接。：/

这是我尝试使用的 "texttoSpeechWidget" 的代码。

using UnityEngine;
using UnityEngine.UI;
using IBM.Watson.DeveloperCloud.Services.TextToSpeech.v1;
using IBM.Watson.DeveloperCloud.Logging;
using IBM.Watson.DeveloperCloud.DataTypes;
using System.Collections.Generic;
using IBM.Watson.DeveloperCloud.Utilities;

#pragma warning disable 414

namespace IBM.Watson.DeveloperCloud.Widgets
{
  /// <summary>
  /// TextToSpeech widget class wraps the TextToSpeech serivce.
  /// </summary>
  [RequireComponent(typeof(AudioSource))]
  public class TextToSpeechWidget : Widget
  {
 #region Inputs
[SerializeField]
private Input m_TextInput = new Input("Text", typeof(TextToSpeechData), 
"OnTextInput");
[SerializeField]
private Input m_VoiceInput = new Input("Voice", typeof(VoiceData), 
"OnVoiceSelect");
#endregion

#region Outputs
[SerializeField]
private Output m_Speaking = new Output(typeof(SpeakingStateData), true);
[SerializeField]
private Output m_DisableMic = new Output(typeof(DisableMicData));
[SerializeField]
private Output m_LevelOut = new Output(typeof(LevelData));
#endregion

#region Private Data
TextToSpeech m_TextToSpeech = new TextToSpeech();

[SerializeField, Tooltip("How often to send level out data in seconds.")]
private float m_LevelOutInterval = 0.05f;
[SerializeField]
private float m_LevelOutputModifier = 1.0f;
[SerializeField]
private Button m_TextToSpeechButton = null;
[SerializeField]
private InputField m_Input = null;
[SerializeField]
private Text m_StatusText = null;
[SerializeField]
private VoiceType m_Voice = VoiceType.en_US_Michael;
[SerializeField]
private bool m_UsePost = false;

private AudioSource m_Source = null;
private int m_LastPlayPos = 0;

private class Speech
{
  ~Speech()
  {
    if (Clip != null)
      UnityObjectUtil.DestroyUnityObject(Clip);
  }

  public bool Ready { get; set; }
  public AudioClip Clip { get; set; }

  public Speech(TextToSpeech textToSpeech, string text, bool usePost)
  {
    textToSpeech.ToSpeech(text, OnAudioClip, usePost);
  }

  private void OnAudioClip(AudioClip clip)
  {
    Clip = clip;
    Ready = true;
  }

};

private Queue<Speech> m_SpeechQueue = new Queue<Speech>();
private Speech m_ActiveSpeech = null;
#endregion

#region Public Memebers

/// <summary>
/// Gets or sets the voice. Default voice is English, US - Michael
/// </summary>
/// <value>The voice.</value>
public VoiceType Voice
{
  get
  {
    return m_Voice;
  }
  set
  {
    m_Voice = value;
  }
}

#endregion

#region Event Handlers
/// <summary>
/// Button event handler.
/// </summary>
public void OnTextToSpeech()
{
  if (m_TextToSpeech.Voice != m_Voice)
    m_TextToSpeech.Voice = m_Voice;
  if (m_Input != null)
    m_SpeechQueue.Enqueue(new Speech(m_TextToSpeech, m_Input.text, m_UsePost));
  if (m_StatusText != null)
    m_StatusText.text = "THINKING";
  if (m_TextToSpeechButton != null)
    m_TextToSpeechButton.interactable = false;
}
#endregion

#region Private Functions
private void OnTextInput(Data data)
{
  TextToSpeechData text = data as TextToSpeechData;
  if (text == null)
    throw new WatsonException("Wrong data type received.");

  if (!string.IsNullOrEmpty(text.Text))
  {
    if (m_TextToSpeech.Voice != m_Voice)
      m_TextToSpeech.Voice = m_Voice;

    m_SpeechQueue.Enqueue(new Speech(m_TextToSpeech, text.Text, m_UsePost));
  }
}

private void OnVoiceSelect(Data data)
{
  VoiceData voice = data as VoiceData;
  if (voice == null)
    throw new WatsonException("Unexpected data type");

  m_Voice = voice.Voice;
}

private void OnEnable()
{
  UnityObjectUtil.StartDestroyQueue();

  if (m_StatusText != null)
    m_StatusText.text = "READY";
}

/// <exclude />
protected override void Start()
{
  base.Start();
  m_Source = GetComponent<AudioSource>();
}

private void Update()
{
  if (m_Source != null && !m_Source.isPlaying
      && m_SpeechQueue.Count > 0
      && m_SpeechQueue.Peek().Ready)
  {
    CancelInvoke("OnEndSpeech");

    m_ActiveSpeech = m_SpeechQueue.Dequeue();
    if (m_ActiveSpeech.Clip != null)
    {
      if (m_Speaking.IsConnected)
        m_Speaking.SendData(new SpeakingStateData(true));
      if (m_DisableMic.IsConnected)
        m_DisableMic.SendData(new DisableMicData(true));

      m_Source.spatialBlend = 0.0f;     // 2D sound
      m_Source.loop = false;            // do not loop
      m_Source.clip = m_ActiveSpeech.Clip;             // clip
      m_Source.Play();

      Invoke("OnEndSpeech", ((float)m_ActiveSpeech.Clip.samples / (float)m_ActiveSpeech.Clip.frequency) + 0.1f);
      if (m_LevelOut.IsConnected)
      {
        m_LastPlayPos = 0;
        InvokeRepeating("OnLevelOut", m_LevelOutInterval, m_LevelOutInterval);
      }
    }
    else
    {
      Log.Warning("TextToSpeechWidget", "Skipping null AudioClip");
    }
  }

  if (m_TextToSpeechButton != null)
    m_TextToSpeechButton.interactable = true;
  if (m_StatusText != null)
    m_StatusText.text = "READY";
}

private void OnLevelOut()
{
  if (m_Source != null && m_Source.isPlaying)
  {
    int currentPos = m_Source.timeSamples;
    if (currentPos > m_LastPlayPos)
    {
      float[] samples = new float[currentPos - m_LastPlayPos];
      m_Source.clip.GetData(samples, m_LastPlayPos);
      m_LevelOut.SendData(new LevelData(Mathf.Max(samples) * m_LevelOutputModifier, m_LevelOutputModifier));
      m_LastPlayPos = currentPos;
    }
  }
  else
    CancelInvoke("OnLevelOut");
}
private void OnEndSpeech()
{
  if (m_Speaking.IsConnected)
    m_Speaking.SendData(new SpeakingStateData(false));
  if (m_DisableMic.IsConnected)
    m_DisableMic.SendData(new DisableMicData(false));
  if (m_Source.isPlaying)
    m_Source.Stop();

  m_ActiveSpeech = null;
}

/// <exclude />
protected override string GetName()
{
  return "TextToSpeech";
}
#endregion

}

Answer 1

首先，您需要编写一些代码来实现它。

使用 Widget class 作为基础创建一个新的 widget。

将您的输入作为 SpeechToText 并将输出作为 TextToSpeech


[SerializeField]
private Input m_SpeechInput = new Input("Text", typeof(SpeechToTextData), "OnSpeech");
[SerializeField]
private Output m_SpeechOutput = new Output(typeof(TextToSpeechData), true);
</pre>

然后您应该将 OnSpeech 函数添加到您的小部件中，以通过创建新的 TextToSpeechData 对象类型并使用输入 SpeechToTextData 的文本（如果它是最终的）来处理从 SpeechToTextData 到 TextToSpeechData 的对话。

像这样在 OnSpeech 中使用您的输出；


if (m_SpeechOutput.IsConnected)
    m_SpeechOutput.SendData( /* object of type  SpeechToTextData*/ );
</pre>

确保场景中存在 SpeechToTextWidget 和 TextToSpeechWidget，并且它们都连接到这个新的小部件。

(Unity 5.6) (IBM Watson SDK) 将用户语音解析为文本，然后再解析为语音

(Unity 5.6) (IBM Watson SDK) Parsing user speech into text and then back into speech

unity3d

sdk

speech-to-text

watson