/// <summary>
/// 获取语法的ID
/// </summary>
/// <returns></returns>
private string GetGrammarID()
{
int ret = -1;
string temp = Application.dataPath;
#if UNITY_ANDROID
temp = Application.persistentDataPath + "/gm_continuous_digit.abnf";
#elif UNITY_STANDALONE_WIN
temp = Application.streamingAssetsPath + "/gm_continuous_digit.abnf";
#endif
byte[] grammarBytes = Utils.ReadFile(temp);
int grammar_len = grammarBytes.Length;
IntPtr ptrGrammarID = DllImports.MSPUploadData("usergram", grammarBytes, (uint)grammar_len, "dtt = abnf, sub = asr", out ret);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
Utils.CustomPrint("MSPUploadData error, errCode=" + ((ErrorCode)ret).ToString("G"));
return(null);
}
string grammarID = Marshal.PtrToStringAnsi(ptrGrammarID);
Utils.CustomPrint("grammarID:" + grammarID + "\r\n");
return(grammarID);
}
/// <summary>
/// 处理错误
/// </summary>
/// <param name="errorMsg">错误的函数</param>
/// <param name="ret">错误的返回值</param>
/// <param name="sessionID">sessionID</param>
/// <param name="fs">打开的文件</param>
private void HandleErrorMsg(string errorMsg, int ret, string sessionID, FileStream fs)
{
Utils.CustomPrint(errorMsg + " err,errCode=" + ((ErrorCode)ret).ToString("G"));
DllImports.QISRSessionEnd(sessionID, errorMsg);
if (fs != null)
{
fs.Close();
fs = null;
}
}
/// <summary>
/// 处理错误
/// </summary>
/// <param name="errorMsg">错误的函数</param>
/// <param name="ret">错误的返回值</param>
/// <param name="sessionID">sessionID</param>
/// <param name="writer">打开的文件</param>
/// <param name="fs">打开的文件</param>
private static void HandleErrorMsg(string errorMsg, int ret, string sessionID, BinaryWriter writer, FileStream fs)
{
Utils.CustomPrint(errorMsg + " err,errCode=" + ((ErrorCode)ret).ToString("G"));
DllImports.QTTSSessionEnd(sessionID, errorMsg);
if (writer != null)
{
writer.Close();
writer = null;
}
if (fs != null)
{
fs.Close();
fs = null;
}
}
/// <summary>
/// 运行语音合成功能
/// </summary>
/// <returns></returns>
public IEnumerator RunTTS()
{
if (text == null)
{
yield break;
}
WaveFormat waveFormat = new WaveFormat();
int nLength = 0;
int nFormatChunkLength = 0x10; // Format chunk length.
short shPad = 1; // File padding
int ret = 0;
uint audioLen = 0;
int sampleCount = 0;
int synthStatus = (int)TTSStatus.MSP_TTS_FLAG_STILL_HAVE_DATA;
IntPtr ptrSessionID = DllImports.QTTSSessionBegin(sessionTTSBeginParams, out ret);
string sessionID = Marshal.PtrToStringAnsi(ptrSessionID);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
Utils.CustomPrint("TTSSessionBegin failed, errCode=" + ((ErrorCode)ret).ToString("G"));
yield break;
}
FileStream waveFile = new FileStream(textPath, FileMode.OpenOrCreate);
BinaryWriter writer = new BinaryWriter(waveFile);
char[] chunkRiff = { 'R', 'I', 'F', 'F' };
char[] chunkType = { 'W', 'A', 'V', 'E' };
char[] chunkFmt = { 'f', 'm', 't', ' ' };
char[] chunkData = { 'd', 'a', 't', 'a' };
//RIFF块
writer.Write(chunkRiff);
writer.Write(nLength);
writer.Write(chunkType);
//WAVE块
writer.Write(chunkFmt);
writer.Write(nFormatChunkLength);
writer.Write(shPad);
writer.Write(waveFormat.mChannels);
writer.Write(waveFormat.mSamplesPerSecond);
writer.Write(waveFormat.mAverageBytesPerSecond);
writer.Write(waveFormat.mBlockAlign);
writer.Write(waveFormat.mBitsPerSample);
//数据块
writer.Write(chunkData);
writer.Write((int)0); // The sample length will be written in later.
ret = DllImports.QTTSTextPut(sessionID, text, (uint)text.Length, null);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QTTSTextPut", ret, sessionID, writer, waveFile);
yield break;
}
Utils.CustomPrint("正在合成 ...\n");
while (true)
{
///获取合成音频
IntPtr data = DllImports.QTTSAudioGet(sessionID, ref audioLen, ref synthStatus, out ret);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
break;
}
if (IntPtr.Zero != data)
{
byte[] buff = new byte[audioLen];
Marshal.Copy(data, buff, 0, buff.Length);
writer.Write(buff, 0, buff.Length);
sampleCount += (int)audioLen;///计算data_size大小
buff = null;
}
if (TTSStatus.MSP_TTS_FLAG_DATA_END == (TTSStatus)synthStatus)
{
break;
}
Utils.CustomPrint(">");
yield return(new WaitForSeconds(0.2f));
}
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
Utils.CustomPrint("QTTSAudioGet failed, errCode=" + ((ErrorCode)ret).ToString("G"));
DllImports.QTTSSessionEnd(sessionID, "AudioGetError");
writer.Close();
waveFile.Close();
yield break;
}
//写WAV文件尾
writer.Seek(4, SeekOrigin.Begin);
writer.Write((int)(sampleCount + 36));
writer.Seek(40, SeekOrigin.Begin);
writer.Write(sampleCount);
writer.Close();
waveFile.Close();
writer = null;
waveFile = null;
///合成完毕
ret = DllImports.QTTSSessionEnd(sessionID, "Normal");
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
Utils.CustomPrint("QTTSSessionEnd failed, errCode=" + ((ErrorCode)ret).ToString("G"));
}
ptrSessionID = IntPtr.Zero;
Utils.CustomPrint("合成完毕");
yield break;
}
/// <summary>
/// 运行声音识别功能
/// </summary>
/// <returns></returns>
public IEnumerator RunISR()
{
///模拟录音，输入音频
if (!File.Exists(path))
{
Utils.CustomPrint("文件" + path + "不存在！");
yield break;
}
const int BUFFER_NUM = 640 * 10;/// 每次写入200ms音频(16k，16bit)：1帧音频20ms，10帧=200ms。16k采样率的16位音频，一帧的大小为640Byte
int ret = 0;
string result = "";
int len;
int audStatus = (int)AudioStatus.MSP_AUDIO_SAMPLE_CONTINUE; ///音频状态
int epStatus = (int)EpStatus.MSP_EP_NULL; ///端点检测
int recStatus = (int)RecogStatus.MSP_REC_NULL; ///识别状态
int rsltStatus = (int)RecogStatus.MSP_REC_NULL;
bool first = true;
int errcode = 0;
Utils.CustomPrint("上传语法 ...+\r\n");
string grammarID = GetGrammarID();
if (!string.IsNullOrEmpty(grammarID))
{
Utils.CustomPrint("上传语法成功\r\n");
}
else
{
Utils.CustomPrint("MSPUploadData error");
yield break;
}
IntPtr ptrSessionID = DllImports.QISRSessionBegin(grammarID, sessionISRBeginParams, out errcode);
string sessionID = Marshal.PtrToStringAnsi(ptrSessionID);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)errcode)
{
Utils.CustomPrint("QISRSessionBegin err,errCode=" + ((ErrorCode)errcode).ToString("G"));
yield break;
}
FileStream fp = new FileStream(path, FileMode.Open);
byte[] buff = new byte[BUFFER_NUM];
IntPtr bp = Marshal.AllocHGlobal(BUFFER_NUM);
while (fp.Position != fp.Length)
{
len = fp.Read(buff, 0, BUFFER_NUM);
Marshal.Copy(buff, 0, bp, buff.Length);
audStatus = (int)AudioStatus.MSP_AUDIO_SAMPLE_CONTINUE;
if (first)
{
audStatus = (int)AudioStatus.MSP_AUDIO_SAMPLE_FIRST;
first = false;
}
Utils.CustomPrint(">");
///开始向服务器发送音频数据
ret = DllImports.QISRAudioWrite(sessionID, bp, (uint)len, audStatus, ref epStatus, ref recStatus);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QISRAudioWrite", ret, sessionID, fp);
yield break;
}
if (EpStatus.MSP_EP_AFTER_SPEECH == (EpStatus)epStatus)
{
break;
}
///sleep一下很有必要，防止MSC端无缓冲的识别结果时浪费CPU资源
yield return(new WaitForSeconds(0.2f));
}
buff = null;
fp.Close();
fp = null;
Marshal.FreeHGlobal(bp);
ret = DllImports.QISRAudioWrite(sessionID, IntPtr.Zero, 0, (int)AudioStatus.MSP_AUDIO_SAMPLE_LAST, ref epStatus, ref recStatus);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QISRAudioWrite", ret, sessionID, null);
yield break;
}
///最后一块数据发完之后，循环从服务器端获取结果
///考虑到网络环境不好的情况下，需要对循环次数作限定
while (RecogStatus.MSP_REC_STATUS_COMPLETE != (RecogStatus)rsltStatus)
{
IntPtr rec_result = DllImports.QISRGetResult(sessionID, ref rsltStatus, 0, out ret);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QISRGetResult", ret, sessionID, null);
yield break;
}
if (IntPtr.Zero != rec_result)
{
string tmp = Marshal.PtrToStringAnsi(rec_result);
Utils.CustomPrint("---:" + tmp);
result += tmp;
Utils.CustomPrint("传完音频后返回结果！:" + result);
}
yield return(new WaitForSeconds(0.2f));
}
ret = DllImports.QISRSessionEnd(sessionID, "Normal");
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
Utils.CustomPrint("QISRSessionEnd failed, errCode=" + ((ErrorCode)ret).ToString("G"));
}
ptrSessionID = IntPtr.Zero;
Utils.CustomPrint("识别完成\r\n");
yield break;
}
/// <summary>
/// 运行语音评测功能
/// </summary>
/// <returns></returns>
public IEnumerator RunISE()
{
int len;
int audStatus = (int)AudioStatus.MSP_AUDIO_SAMPLE_CONTINUE; ///音频状态
int epStatus = (int)EpStatus.MSP_EP_NULL; ///端点检测
int recStatus = (int)RecogStatus.MSP_REC_NULL; ///识别状态
bool first = true;
uint rlstLen = 0;
IntPtr rsltPrt = IntPtr.Zero;
int ret = 0;
const int BUFFER_NUM = 640 * 10;/// 每次写入200ms音频(16k，16bit)：1帧音频20ms，10帧=200ms。16k采样率的16位音频，一帧的大小为640Byte
FileStream textFS = null;
FileStream waveFS = null;
string parameters, textPath, wavePath;
GetValuesByCategory(category, out parameters, out textPath, out wavePath);
#if UNITY_ANDROID
textPath = Application.persistentDataPath + "/" + textPath;
wavePath = Application.persistentDataPath + "/" + wavePath;
#elif UNITY_STANDALONE_WIN
textPath = Application.streamingAssetsPath + "/" + textPath;
wavePath = Application.streamingAssetsPath + "/" + wavePath;
#endif
IntPtr ptrSessionID = DllImports.QISESessionBegin(parameters, "", out ret);
string sessionID = Marshal.PtrToStringAnsi(ptrSessionID);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
Utils.CustomPrint("QISESessionBegin err,errCode=" + ((ErrorCode)ret).ToString("G"));
yield break;
}
if (!File.Exists(textPath) || !File.Exists(wavePath))
{
Utils.CustomPrint("文件" + textPath + " 或者 " + wavePath + "不存在！");
DllImports.QISESessionEnd(sessionID, "File not exist");
yield break;
}
textFS = new FileStream(textPath, FileMode.Open);
int textLength = (int)textFS.Length;
byte[] buff = new byte[textLength];
textFS.Read(buff, 0, textLength);
textFS.Close();
textFS = null;
ret = DllImports.QISETextPut(sessionID, Encoding.Default.GetString(buff), (uint)textLength, "");
Utils.CustomPrint("Encoding:" + Encoding.Default.EncodingName.ToLower());
Utils.CustomPrint(Regex.Unescape(Encoding.Default.GetString(buff)) + "|length:" + textLength);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QISETextPut", ret, sessionID, null);
yield break;
}
waveFS = new FileStream(wavePath, FileMode.Open);
buff = new byte[BUFFER_NUM];
IntPtr bp = Marshal.AllocHGlobal(BUFFER_NUM);
while (waveFS.Position != waveFS.Length)
{
len = waveFS.Read(buff, 0, BUFFER_NUM);
Marshal.Copy(buff, 0, bp, buff.Length);
audStatus = (int)AudioStatus.MSP_AUDIO_SAMPLE_CONTINUE;
if (first)
{
audStatus = (int)AudioStatus.MSP_AUDIO_SAMPLE_FIRST;
first = false;
}
Utils.CustomPrint(">" + len);
///开始向服务器发送音频数据
ret = DllImports.QISEAudioWrite(sessionID, bp, (uint)len, audStatus, ref epStatus, ref recStatus);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QISEAudioWrite", ret, sessionID, waveFS);
yield break;
}
if (RecogStatus.MSP_REC_STATUS_SUCCESS == (RecogStatus)recStatus)
{
rsltPrt = DllImports.QISEGetResult(sessionID, ref rlstLen, ref recStatus, ref ret);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QISEGetResult", ret, sessionID, waveFS);
yield break;
}
if (IntPtr.Zero != rsltPrt)
{
Utils.CustomPrint("rlstLen:" + rlstLen);
byte[] data = new byte[rlstLen];
Marshal.Copy(rsltPrt, data, 0, (int)rlstLen);
Utils.CustomPrint("+++传完音频后返回结果！:" + GBToUnicode(data, data.Length));
}
}
if (EpStatus.MSP_EP_AFTER_SPEECH == (EpStatus)epStatus)
{
break;
}
///sleep一下很有必要，防止MSC端无缓冲的识别结果时浪费CPU资源
yield return(new WaitForSeconds(0.2f));
}
buff = null;
waveFS.Close();
waveFS = null;
Marshal.FreeHGlobal(bp);
ret = DllImports.QISEAudioWrite(sessionID, IntPtr.Zero, 0, (int)AudioStatus.MSP_AUDIO_SAMPLE_LAST, ref epStatus, ref recStatus);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QISEAudioWrite", ret, sessionID, null);
yield break;
}
while (RecogStatus.MSP_REC_STATUS_COMPLETE != (RecogStatus)recStatus)
{
rsltPrt = DllImports.QISEGetResult(sessionID, ref rlstLen, ref recStatus, ref ret);
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
HandleErrorMsg("QISEGetResult", ret, sessionID, null);
yield break;
}
if (IntPtr.Zero != rsltPrt)
{
Utils.CustomPrint("rlstLen:" + rlstLen);
byte[] data = new byte[rlstLen];
Marshal.Copy(rsltPrt, data, 0, (int)rlstLen);
Utils.CustomPrint("---传完音频后返回结果！:" + GBToUnicode(data, data.Length));
}
yield return(new WaitForSeconds(0.2f));
}
rsltPrt = IntPtr.Zero;
ret = DllImports.QISESessionEnd(sessionID, "Normal");
if (ErrorCode.MSP_SUCCESS != (ErrorCode)ret)
{
Utils.CustomPrint("QISESessionEnd failed, errCode=" + ((ErrorCode)ret).ToString("G"));
}
ptrSessionID = IntPtr.Zero;
Utils.CustomPrint("评测完成\r\n");
yield break;
}