/// <summary>
/// 提取识别结果
/// </summary>
/// <param name="result">识别结果结构体</param>
private void GetAsrResult(ASR_RECOG_RESULT result)
{
int sizeOfItem = Marshal.SizeOf(typeof(ASR_RECOG_RESULT_ITEM));
for (int i = 0; i < (int)result.uiResultItemCount; ++i)
{
ASR_RECOG_RESULT_ITEM item = (ASR_RECOG_RESULT_ITEM)Marshal.PtrToStructure((IntPtr)(result.psResultItemList.ToInt32() + i * sizeOfItem), typeof(ASR_RECOG_RESULT_ITEM));
int len = FindNullIndex(item.pszResult, 100);
byte[] data = new byte[len];
Marshal.Copy(item.pszResult, data, 0, len);
_sbResult.Append(Encoding.UTF8.GetString(data));
}
}
public static extern int hci_asr_free_recog_result(ref ASR_RECOG_RESULT recogResult);
public static extern int hci_asr_recog(int sessionId, byte[] data, uint dataLen, string congfig, string grammarData, ref ASR_RECOG_RESULT recogResult);
/// <summary>
/// 语音识别
/// </summary>
/// <param name="audioData">音频数据</param>
/// <param name="sessionId">会话ID</param>
/// <returns>true-成功；false-失败</returns>
private bool Recog(byte[] audioData, int sessionId)
{
int errCode = 0;
bool success = false;
int perLen = 6400; // 0.2s
int len = 0; // 已传入的长度
ASR_RECOG_RESULT result = new ASR_RECOG_RESULT();
// 实时识别过程模拟，将待识别的音频数据分为多段，逐段传入识别接口
// 若某次传输音频数据检测到末端，则跳出循环，终止音频数据传入，以准备获取识别结果
while (len < audioData.Length)
{
// 本次要传入的参与识别的数据长度
// 剩余的多余 perLen 则传入 perLen 个字节，若不足则传入剩余数据
int currLen = 0;
if (audioData.Length - len >= perLen)
{
currLen = perLen;
}
else
{
currLen = audioData.Length - len;
}
byte[] currData = new byte[currLen];
Buffer.BlockCopy(audioData, len, currData, 0, currLen);
errCode = hci_api.hci_asr_recog(sessionId, currData, (uint)currLen, null, null, ref result);
if (result.uiResultItemCount > 0)
{
GetAsrResult(result);
hci_api.hci_asr_free_recog_result(ref result);
}
if (errCode == (int)HCI_ERR_CODE.HCI_ERR_ASR_REALTIME_END)
{
errCode = hci_api.hci_asr_recog(sessionId, null, 0, null, null, ref result);
if (errCode == (int)HCI_ERR_CODE.HCI_ERR_NONE)
{
GetAsrResult(result);
hci_api.hci_asr_free_recog_result(ref result);
}
else
{
_errMsg = string.Format("hci_asr_recog return {0}:{1}", errCode, hci_api.hci_get_error_info(errCode));
success = false;
break;
}
}
else if (errCode == (int)HCI_ERR_CODE.HCI_ERR_ASR_REALTIME_WAITING ||
errCode == (int)HCI_ERR_CODE.HCI_ERR_ASR_REALTIME_NO_VOICE_INPUT)
{
// 在连续识别的场景，忽略这两个情况，继续识别后面的音频
len += currLen;
}
else
{
// 识别失败
_errMsg = string.Format("hci_asr_recog return {0}:{1}", errCode, hci_api.hci_get_error_info(errCode));
success = false;
break;
}
// TODO:需测试是否需要 Sleep() 方法以提高识别速度
// Thread.Sleep(200); // 模拟真实说话人语速，发送200ms数据后需等待200ms
}
// 若检测到端点，但数据已经传入完毕，则需要告诉引擎数据输入完毕
// 或者检测到末端了，也需要告诉引擎，获取结果
if (errCode == (int)HCI_ERR_CODE.HCI_ERR_ASR_REALTIME_WAITING ||
errCode == (int)HCI_ERR_CODE.HCI_ERR_ASR_REALTIME_END)
{
errCode = hci_api.hci_asr_recog(sessionId, null, 0, null, null, ref result);
if (errCode == (int)HCI_ERR_CODE.HCI_ERR_NONE)
{
success = true;
GetAsrResult(result);
hci_api.hci_asr_free_recog_result(ref result);
}
else
{
success = false;
_errMsg = string.Format("hci_asr_recog return {0}:{1}", errCode, hci_api.hci_get_error_info(errCode));
}
}
return(success);
}
