Unity 语音转文字

1. Window 自带的微软语音识别

参考官方手册：Windows.Speech.DictationRecognizer - Unity 脚本 API (unity3d.com)

这个中英识别都比较准确，不过应该只支持Win10以上。

需在设置里开启在线语音识别。

 代码部分：

using System;
using System.Collections;
using System.Collections.Generic;
using UnityEngine;
using UnityEngine.Events;
using UnityEngine.UI;
using UnityEngine.Windows.Speech;

public class DictationScript : MonoBehaviour
{
    public Text hypothesesText;//显示输入过程中猜想结果的
    public Text recognitionsText;//显示识别结果的
    public Button startBtn;
    public Button stopBtn;

    private DictationRecognizer dictationRecognizer;

    void Start()
    {
        dictationRecognizer = new DictationRecognizer();

        dictationRecognizer.DictationResult += OnDictationResult;
        dictationRecognizer.DictationHypothesis += OnDictationHypothesis;
        dictationRecognizer.DictationComplete += OnDictationComplete;
        dictationRecognizer.DictationError += OnDictationError;

        //dictationRecognizer.Start();
        startBtn.onClick.AddListener(DictationStart);
        stopBtn.onClick.AddListener(DictionStop);
    }

    private void OnDestroy()
    {
        dictationRecognizer.Stop();
        dictationRecognizer.Dispose();
    }

    /// <summary>
    /// 语音识别结果
    /// </summary>
    /// <param name="text">识别结果</param>
    /// <param name="confidence"></param>
    private void OnDictationResult(string text, ConfidenceLevel confidence)
    {
        Debug.LogFormat("识别结果: {0}", text);
        recognitionsText.text = text;

        DictionStop();//我是希望得到结果就自动停止输入
    }

    /// <summary>
    /// 语音输入过程中对结果猜想时触发的事件。
    /// </summary>
    /// <param name="text">识别猜想</param>
    private void OnDictationHypothesis(string text)
    {
        Debug.LogFormat("识别猜想: {0}", text);
        hypothesesText.text = text;
    }

    private void OnDictationComplete(DictationCompletionCause cause)
    {
        if (cause != DictationCompletionCause.Complete)
            Debug.LogErrorFormat("识别失败: {0}.", cause);
    }

    private void OnDictationError(string error, int hresult)
    {
        Debug.LogErrorFormat("识别错误: {0}; HResult = {1}.", error, hresult);
    }

    /// <summary>
    /// 开启听写识别会话
    /// </summary>
    public void DictationStart()
    {
        dictationRecognizer.Start();
        startBtn.gameObject.SetActive(false);
        stopBtn.gameObject.SetActive(true);
    }

    /// <summary>
    /// 结束听写识别会话
    /// </summary>
    public void DictionStop()
    {
        dictationRecognizer.Stop();
        startBtn.gameObject.SetActive(true);
        stopBtn.gameObject.SetActive(false);
    }

}

2. 百度SDK

也可以考虑其他平台的SDK，我只测试了百度的，这个英文识别不大好（最起码Unity识别不出来，IPad可以识别出来），上面那个中英识别都不错，不过只支持Window，百度的这个我也没测试其他平台，自行测试吧。

官方文档：https://ai.baidu.com/ai-doc/SPEECH/ilbxfvpau

1. 在百度智能云控制台创建一个应用，勾上短语音识别
2. 下载C# SDK：https://ai.baidu.com/sdk#asr
3. 将SDK 放入Unity中Plugins文件夹下

然后直接上代码吧

using System.Collections;
using System.Collections.Generic;
using UnityEngine;
using Baidu.Aip.Speech;
using UnityEngine.UI;

public class BaiduSpeech : MonoBehaviour
{
    public Text text;//显示结果用的

    // 设置APPID/AK/SK
    string APP_ID = "你的 AppID";
    string API_KEY = "你的 API Key";
    string SECRET_KEY = "你的 Secret Key";

    int rate = 8000;

    Asr client;
    AudioSource aud;
    int audioLength;//录音的长度

    private void Awake()
    {
        if (GetComponent<AudioSource>() == null)
            aud = gameObject.AddComponent<AudioSource>();
        else
            aud = gameObject.GetComponent<AudioSource>();
        aud.playOnAwake = false;
    }
    // Start is called before the first frame update
    void Start()
    {
        client = new Asr(APP_ID, API_KEY, SECRET_KEY);
        client.Timeout = 60000;  // 修改超时时间
    }

    /// <summary>
    /// 开始录音
    /// </summary>
    public void StartMic()
    {
        if (Microphone.devices.Length == 0) return;
        Microphone.End(null);
        Debug.Log("Start");
        aud.clip = Microphone.Start(null, false, 10, rate);
    }

    /// <summary>
    /// 结束录音
    /// </summary>
    public void EndMic()
    {
        int lastPos = Microphone.GetPosition(null);
        if (Microphone.IsRecording(null))
            audioLength = lastPos / rate;//录音时长  
        else
            audioLength = 10;
        Debug.Log("Stop");
        Microphone.End(null);

        AsrData(GetClipData());
    }

    /// <summary>
    /// 把录音转换为Byte[]
    /// </summary>
    /// <returns></returns>
    public byte[] GetClipData()
    {
        if (aud.clip == null)
        {
            Debug.LogError("录音数据为空");
            return null;
        }

        float[] samples = new float[aud.clip.samples];
        aud.clip.GetData(samples, 0);

        byte[] outData = new byte[samples.Length * 2];
        int rescaleFactor = 32767; //to convert float to Int16   
        for (int i = 0; i < samples.Length; i++)
        {
            short temshort = (short)(samples[i] * rescaleFactor);

            byte[] temdata = System.BitConverter.GetBytes(temshort);

            outData[i * 2] = temdata[0];
            outData[i * 2 + 1] = temdata[1];
        }
        if (outData == null || outData.Length <= 0)
        {
            Debug.LogError("录音数据为空");
            return null;
        }

        return outData;
    }

    // 识别本地文件
    public void AsrData(byte[] data)
    {
        // 可选参数
        var options = new Dictionary<string, object>
        {
            {"dev_pid", 1537}
        };
        client.Timeout = 120000; // 若语音较长，建议设置更大的超时时间. ms
        var result = client.Recognize(data, "pcm", rate, options);
        if (result["result"] != null)
        {
            text.text = result["result"][0].ToString();
        }
        else
        {
            Debug.Log("识别错误");
        }
        Debug.Log(result);
    }

}