/// <summary>
/// 语音识别。如果是标准音频格式：pcm/16k/16位/单通道，则调用此方法。如果是其他格式的音频，请调用另一个方法并传入音频格式 WaveFormat 参数。
/// </summary>
/// <param name="audioData">小于 60s 的音频数据（总长度不超过1920k），音频格式要求：pcm/16k/16位/单通道 。</param>
/// <param name="languageType">音频语种类型</param>
/// <param name="recogResult">识别成功返回识别结果，识别失败返回错误消息</param>
/// <returns>识别成功或失败，true-成功；false-失败</returns>
public bool AudioRecog(byte[] audioData, LanguageType languageType, out string recogResult)
{
if (_asr == null)
{
recogResult = "客户端尚未初始化";
return(false);
}
return(_asr.AudioRecog(audioData, languageType, out recogResult));
}
private void btnRecog_Click(object sender, EventArgs e)
{
string fileName = txtAudioFile.Text.Trim();
if (string.IsNullOrEmpty(fileName))
{
return;
}
if (!File.Exists(fileName))
{
return;
}
btnRecog.Enabled = false;
string text = cmbLanguage.Text;
ThreadPool.QueueUserWorkItem(o =>
{
WriteLine("Start ...");
string result = string.Empty;
byte[] data = File.ReadAllBytes(fileName);
LanguageType type = GetLanguageType(text);
Stopwatch sw = new Stopwatch();
sw.Start();
if (cbPCM.Checked)
{
// 3）语音识别
_asr.AudioRecog(data, type, out result);
}
else
{
WaveFormat waveFormat = WaveFormat.PCM_MONO;
waveFormat.nSamplesPerSec = 22050; // 默认值就是这样的
waveFormat.nAvgBytesPerSec = 22050 * 2;
_asr.AudioRecog(data, type, out result, waveFormat);
}
sw.Stop();
WriteLine(result);
WriteLine("Finish! 耗时：" + sw.ElapsedMilliseconds + " 毫秒。\n");
EnableBtn(true);
});
}
private void Asr(List <string> wavList)
{
for (int i = 0; i < wavList.Count; i++)
{
string wavFileName = wavList[i];
string pcmFileName = wavFileName.Replace(".wav", ".pcm");
if (!File.Exists(pcmFileName))
{
continue;
}
byte[] data = File.ReadAllBytes(pcmFileName);
LanguageType languageType = LanguageType.Mandarin;
string result = string.Empty;
bool ret = _asr.AudioRecog(data, languageType, out result);
if (ret)
{
AddNewData(wavFileName, result);
}
UpdateProcessBar(i + 1);
}
UpdateProcessBar(_progressBar1.Maximum);
MessageBox.Show("音频文件识别完成！");
_isBusy = false;
}
