IEnumerator Recognition(AudioClip clip2Send)
{
float[] sample = new float[recordClip.samples];
recordClip.GetData(sample, 0);
short[] intData = new short[sample.Length];
byte[] byteData = new byte[intData.Length * 2];
for (int i = 0; i < sample.Length; i++)
{
intData[i] = (short)(sample[i] * short.MaxValue);
}
Buffer.BlockCopy(intData, 0, byteData, 0, byteData.Length);
var result = aipClient.Recognize(byteData, "pcm", 16000);
var speaking = result.GetValue("result");
if (speaking == null)
{
SetRecognizeText(TipsReference.NOTHING_RECORD);
StopAllCoroutines();
yield return(null);
}
string usefulText = speaking.First.ToString();
SetRecognizeText(usefulText);
yield return(0);
}
// 识别本地文件
public string AsrData(string fileFullName)
{
var data = File.ReadAllBytes(fileFullName);
var result = _asrClient.Recognize(data, "pcm", 16000);
return(result.ToString());
}
// 识别本地文件
public string AsrData(string pcmFilePath)
{
var data = File.ReadAllBytes(pcmFilePath);
var result = _asrClient.Recognize(data, "pcm", 16000);
return(result.ToString());
}
private void OnClickStopButton()
{
StartButton.gameObject.SetActive(false);
StopButton.gameObject.SetActive(false);
DescriptionText.text = "转换中...";
Microphone.End(null); //清空
var data = Asr.ConvertAudioClipToPCM16(_clipRecord); //转换成byte数据
//将byte传入 然后转换
StartCoroutine(_asr.Recognize(data, s =>
{
//转换回掉函数 文本=转换结果不为空且结果长度大于0 否则为“未识别声音”
if (s.result == null)
{
DescriptionText.text = "未识别到声音";
StartButton.gameObject.SetActive(true);
}
else
{
RecognitionResult(s.result[0]);
}
}));
}
public void StopAndGetResult()
{
string result = "";
Microphone.End(null);
var data = Asr.ConvertAudioClipToPCM16(_clipRecord);
StartCoroutine(_asr.Recognize(data, s =>
{
result = s.result != null && s.result.Length > 0 ? s.result[0] : "未识别到声音";
lastContent = result;
if (Eve解析完成 != null && result != "未识别到声音")
{
Eve解析完成(result);
}
else if (Eve未检测到声音 != null && result == "未识别到声音")
{
Eve未检测到声音();
}
//Debug.Log("结束录音：" + result);
}));
if (Eve结束录音 != null)
{
Eve结束录音();
}
}
// 识别本地文件
public void AsrData()
{
var data = File.ReadAllBytes("语音pcm文件地址");
var result = _asrClient.Recognize(data, "pcm", 16000);
Console.Write(result);
}
// 识别本地文件
static public void AsrData()
{
var data = File.ReadAllBytes(@"H:\yank\企业管理系统V1.0\Web\AI\test\16k.wav");
var result = client.Recognize(data, "pcm", 16000);
Console.Write(result);
}
//语音识别
public void recoVideo()
{
Thread.Sleep(2000);
if (client == null)
{
var API_KEY = "uz804M95AKkq5OffLCB9GD3q";
var SECRET_KEY = "IZxCofo7vFRYiPKhXqdmh2Da51MrhDKA";
client = new Baidu.Aip.Speech.Asr(API_KEY, SECRET_KEY);
}
for (int i = 1; i < number; i++)
{
try {
var data = FileUtil.getFileVideo(videoPath, i + ".pcm");
client.Timeout = 12000; // 若语音较长，建议设置更大的超时时间. ms
var result = client.Recognize(data, "pcm", 16000, dic);
if (result["result"] != null)
{
String value = result["result"][0].ToString();
tranText(value, i, captionDic);
}
}
catch (Exception e) {
Console.WriteLine(e.Message);
tranText(e.Message, i, captionDic);
}
}
}
// 识别本地文件
public static AsrResult AsrData(string path)
{
var data = File.ReadAllBytes(path);
var result = AsrClient.Recognize(data, "pcm", 8000);
return(result.ToObject <AsrResult>());
}
// 识别本地文件
public void AsrData(string filePath)
{
var data = File.ReadAllBytes(filePath);
var result = _asrClient.Recognize(data, "pcm", 16000);
Console.Write(result);
}
// 识别本地文件
/// <summary>
///
/// </summary>
/// <param name="filePath">文件地址</param>
/// <param name="fileType">文件类型</param>
/// <param name="rate">采样率</param>
/// <param name="lan">识别语言，默认lan=zh,设置为lan=en后，返回结果一定为英文</param>
public List <string> AsrData(string filePath, string fileType, int rate, string lan)
{
var data = File.ReadAllBytes(filePath);
Dictionary <string, object> para = new Dictionary <string, object>();
para.Add("lan", lan);
if ("pcm".Equals(fileType) || "wav".Equals(fileType))
{
string API_KEY = "PhUlgrvjGmIaSuPAvn5CWbAE";
string SECRET_KEY = "fDZXemDEyPc6dcAfT8YxkgQuN29kVjwC";
Asr asr = new Asr(API_KEY, SECRET_KEY);
JObject result = asr.Recognize(data, fileType, rate, para);
List <string> res = new List <string>();
if ("0".Equals(result["err_no"].ToString()))
{
JArray jaResult = JArray.Parse(result["result"].ToString());
foreach (JToken jt in jaResult)
{
res.Add(jt.ToString());
}
}
return(res);
}
else
{
return(null);
}
}
private void btn_Recognize_Click(object sender, EventArgs e)
{
string fileName = "ffmpeg.exe";
Process p = new Process();
p.StartInfo.UseShellExecute = false;
p.StartInfo.RedirectStandardOutput = true;
p.StartInfo.FileName = fileName;
p.StartInfo.CreateNoWindow = true;
p.StartInfo.Arguments = "-i " + txt_AudioFile.Text + " -y -ar 16000 -ac 1 -acodec pcm_s16le input.wav";//参数以空格分隔，如果某个参数为空，可以传入””
p.Start();
p.WaitForExit();
var data = File.ReadAllBytes("input.wav");
File.Delete("input.wav");
var result = _asrClient.Recognize(data, "pcm", 16000);
Console.Write(result);
if (result["err_msg"].ToString() == "success.")
{
rtb_textOut.Text = result["result"].ToString().Replace("[", "").Replace("]", "").Replace("\"", "").Trim();
rtb_SendMsg.Text = rtb_textOut.Text;
message_bar.Text = "识别成功";
}
else
{
message_bar.Text = "识别失败";
panel_sample.BackColor = Color.Red;
}
}
// 识别本地文件
public void AsrData()
{
var data = File.ReadAllBytes(@"C:\Users\hym\Downloads\aip-speech-csharp-sdk-1.0.0\Speech\re.txt");
var result = _asrClient.Recognize(data, "pcm", 16000);
Console.Write(result);
}
/// <summary>
/// 开始识别按钮
///
/// dev_pid 可选参数
/// 1536 普通话(支持简单的英文识别) 搜索模型 无标点 支持自定义词库 http://vop.baidu.com/server_api
/// 1537 普通话(纯中文识别) 输入法模型 有标点 不支持自定义词库 http://vop.baidu.com/server_api
/// 1737 英语 无标点 不支持自定义词库 http://vop.baidu.com/server_api
/// 1637 粤语 有标点 不支持自定义词库 http://vop.baidu.com/server_api
/// 1837 四川话 有标点 不支持自定义词库 http://vop.baidu.com/server_api
/// 1936 普通话远场 远场模型 有标点 不支持自定义词库 http://vop.baidu.com/server_api
/// 80001语音识别极速版(收费)
/// 80001 普通话 极速版输入法模型 有标点 支持自定义词库 http://vop.baidu.com/pro_api
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
private void button2_Click(object sender, EventArgs e)
{
var type = videoType.Text;
var filePath = this.filePath.Text;
//设置语音识别应用的账号信息(百度智能云管理中心) APP_ID/API_KEY/SECRET_KEY
//demo里的账号
//var APP_ID = "14433392";
//var API_KEY = "C7WMYgLeWv3Wm2yogwv5gD08";
//var SECRET_KEY = "xcvwiwikALBDBaIcGisNQ6aQImtj3qua";
//请更改成自己的账号
string APP_ID = "16982575";
string API_KEY = "0kC4dDwWl3hqo2xz2c4113ZP";
string SECRET_KEY = "8aneCyn9KVAWjGKAIfpcr3vMCFt19kIb";
var client = new Asr(APP_ID, API_KEY, SECRET_KEY);
client.Timeout = 120000; // 若语音较长，建议设置更大的超时时间. ms
// 可选参数
var options = new Dictionary <string, object>
{
{ "dev_pid", 1536 }
};
var result = client.Recognize(File.ReadAllBytes(filePath), type.TrimStart('.'), 16000, options);
voiceResult.Text = Convert.ToString(result);
result.TryGetValue("result", out JToken resultStr);
if (Convert.ToString(resultStr) != "")
{
voiceResult.Text = Convert.ToString(resultStr);
}
}
// 识别本地文件
public void AsrData()
{
var data = System.IO.File.ReadAllBytes(@"C:\Users\Admin\Source\Repos\TTS\AAI.WEB\App_LocalResources\2017年8月8日 16_49_17.wav");
var result = _asrClient.Recognize(data, "wav", 8000);
Response.Write(result);
}
// 识别本地文件
public async Task <string> AsrData(string fileUrl)
{
return
(await Task.Run(() =>
{
var data = File.ReadAllBytes(fileUrl);
var result = _asrClient.Recognize(data, "pcm", 8000);
return JsonConvert.DeserializeObject <string[]>(result["result"].ToString())[0];
}));
}
private JObject Baidu_RecognizeByAPI(string filePhysicsPath, bool isEn)
{
Asr _asrClient = new Asr("TmZSp2x7nZXlPIW6GapK45XG", "BuWaLotBu4ho1GnblaMbue3ppryGuOYG ");
var data = File.ReadAllBytes(filePhysicsPath);
Dictionary <string, object> dic = new Dictionary <string, object>();
dic.Add("lan", "en");
var result = _asrClient.Recognize(data, "WAV", 8000, dic);
return(result);
}
public string Recognition(byte[] data)
{
// 可选参数
var options = new Dictionary <string, object>
{
{ "dev_pid", 1536 }
};
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", 16000, options);
return(result.ToString());
}
public SpeechRecRes SpeechRec(byte[] data)
{
SpeechRecRes res = null;
if (data != null)
{
JObject result = _asr.Recognize(data, _speechFormat, 16000);
res = result?.ToObject <SpeechRecRes>();
}
return(res);
}
/// <summary>
/// 通过文件路径，调用百度语音识别接口返回识别结果
/// </summary>
/// <param name="filePath">文件路径</param>
/// <param name="audioFormat">音频格式</param>
/// <returns></returns>
public static JObject AsrData(string filePath, string audioFormat)
{
try
{
var data = File.ReadAllBytes(filePath);
_asrClient.Timeout = 120000; // 若语音较长，建议设置更大的超时时间. ms
var result = _asrClient.Recognize(data, audioFormat, 16000);
return(result);
}
catch (Exception ex)
{
throw (ex);
}
}
private void OnClickStopButton()
{
StartButton.gameObject.SetActive(false);
StopButton.gameObject.SetActive(false);
DescriptionText.text = "Recognizing...";
Microphone.End(null);
var data = Asr.ConvertAudioClipToPCM16(_clipRecord);
StartCoroutine(_asr.Recognize(data, s =>
{
message = s.result != null && s.result.Length > 0 ? s.result[0] : "未识别到声音";
char[] messageChar = message.ToCharArray();
int temporary = 1;
int chamber = ToDigit(messageChar[(message.IndexOf("病房") - temporary)]);
for (temporary = 1; temporary < message.IndexOf("病房");)
{ //如病房前面还有字 第六个病房 etc. 接着往前找
if (chamber == 0)
{
chamber = ToDigit(message[(message.IndexOf("病房") - (++temporary))]);
}
else if (chamber != 0)
{//找到了
break;
}
}
message2 = s.result != null && s.result.Length > 0 ? s.result[0] : "未识别到声音";
//0none，1开灯2关灯3消毒4通风
int operate = 0;
string[] operates = { "", "开灯", "关灯", "消毒", "通风" };
for (int i = 1; i <= 4;)
{
int t = message2.IndexOf(operates[i]);
if (t == (-1))
{
i++;
}
else
{
operate = i;
break;
}
}//用indexof寻找关键字，没找到返回-1 找到了赋给operate
// chamber 1，2，3，4，5，6，7，8
// operate 1，2，3，4
DescriptionText.text = chamber.ToString() + "," + operate;
StartButton.gameObject.SetActive(true);
}));
}
private void StopRecording()
{
//此时判断是否处于说话状态，如果是说话状态则将按钮复位，并上传说话内容
if (StartSpeech)
{
//上传内容
Log.Debug("麦克风停止录音");
Microphone.End(null);
//转换格式
var data = Asr.ConvertAudioClipToPCM16(clipRecord);
asr.Recognize(data, s =>
{
Log.Debug("进来了");
if (s.result == null && s.result.Length < 0)
{
Log.Debug("结果为空，表示麦克风未识别到声音");
//提示有问题，复位
entry2.callback.AddListener(N);
StartSpeech = false;
}
else
{
//有结果，发送给机器人进行语音回复
tts.Synthesis(s.result[0], r =>
{
if (r.Success)
{
//正常播放
Log.Debug("合成成功，正在播放,声音有几秒：" + audioSource.clip.length);
audioSource.clip = r.clip;
audioSource.Play();
//复位
entry2.callback.AddListener(N);
StartSpeech = false;
}
else
{
//这是出问题了
Debug.Log(s.err_msg);
//提示有问题，复位
entry2.callback.AddListener(N);
StartSpeech = false;
}
});
}
});
}
}
private void OnClickStopButton()
{
StartButton.gameObject.SetActive(false);
StopButton.gameObject.SetActive(false);
DescriptionText.text = "Recognizing...";
Microphone.End(null);
Debug.Log("end record");
var data = Asr.ConvertAudioClipToPCM16(_clipRecord);
StartCoroutine(_asr.Recognize(data, s =>
{
DescriptionText.text = s.result[0];
StartButton.gameObject.SetActive(true);
}));
}
private void OnClickStopButton()
{
StartButton.gameObject.SetActive(false);
StopButton.gameObject.SetActive(false);
DescriptionText.text = "Recognizing...";
Microphone.End(null);
Debug.Log("[WitBaiduAip demo]end record");
var data = Asr.ConvertAudioClipToPCM16(_clipRecord);
ReplayButton.gameObject.SetActive(true);
StartCoroutine(_asr.Recognize(data, s =>
{
DescriptionText.text = s.result != null && s.result.Length > 0 ? s.result[0] : "未识别到声音";
StartButton.gameObject.SetActive(true);
}));
}
public static string GetText(string filename)
{
if (File.Exists(filename) == false)
{
return(null);
}
var data = File.ReadAllBytes(filename);
client.Timeout = 120000; // 若语音较长，建议设置更大的超时时间. ms
Dictionary <string, object> dic = new Dictionary <string, object>();
dic.Add("dev_pid", 1737);
var result = client.Recognize(data, "wav", 16000, dic);
return(result["result"][0].ToString());
}
public static string GetSpeechString(byte[] data)
{
// 设置APPID/AK/SK
string APP_ID = "16787165";
string API_KEY = "k7s2FUxEx2xuVFy6B0VKRH5C";
string SECRET_KEY = "no8gKK05u6Hs2FVCuTGDrXPVYEXM1EpA ";
var client = new Asr(APP_ID, API_KEY, SECRET_KEY);
// 可选参数
var options = new Dictionary <string, object>
{
{ "dev_pid", 1536 }
};
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, "wav", 16000, options).ToObject <BaiduSpeechResult>();
return(result.result.FirstOrDefault());
}
IEnumerator Timer()
{
while (true)
{
yield return(new WaitForSeconds(inteval));
Microphone.End(null);
// print(_clipRecord.length);
updateClips();
AudioClip currentClip = AudioClipHelper.Combine((AudioClip[])_audioClips.ToArray(typeof(AudioClip)));
// AudioClip currentClip = AudioClipHelper.Combine(_audioClips);
// AudioClip currentClip = (AudioClip) _audioClips[0];
var data = Asr.ConvertAudioClipToPCM16(currentClip);
StartCoroutine(_asr.Recognize(data, s =>
{
String text = s.result != null && s.result.Length > 0 ? s.result[0] : "未识别到声音";
Debug.Log(DateTime.Now + "：" + text);
}));
}
}
//语音识别
private void button4_Click(object sender, EventArgs e)
{
//设置语音识别应用的账号信息(百度智能云管理中心) APP_ID/API_KEY/SECRET_KEY
//demo里的账号
//var APP_ID = "14433392";
//var API_KEY = "C7WMYgLeWv3Wm2yogwv5gD08";
//var SECRET_KEY = "xcvwiwikALBDBaIcGisNQ6aQImtj3qua";
//请更改成自己的账号
string APP_ID = "16982575";
string API_KEY = "0kC4dDwWl3hqo2xz2c4113ZP";
string SECRET_KEY = "8aneCyn9KVAWjGKAIfpcr3vMCFt19kIb";
var client = new Asr(APP_ID, API_KEY, SECRET_KEY);
var videoPath = comboBox4.Text;
var type = Path.GetExtension(videoPath);
var data = File.ReadAllBytes(videoPath);
//dev_pid 可选参数
//1536 普通话(支持简单的英文识别) 搜索模型 无标点 支持自定义词库 http://vop.baidu.com/server_api
//1537 普通话(纯中文识别) 输入法模型 有标点 不支持自定义词库 http://vop.baidu.com/server_api
//1737 英语 无标点 不支持自定义词库 http://vop.baidu.com/server_api
//1637 粤语 有标点 不支持自定义词库 http://vop.baidu.com/server_api
//1837 四川话 有标点 不支持自定义词库 http://vop.baidu.com/server_api
//1936 普通话远场 远场模型 有标点 不支持自定义词库 http://vop.baidu.com/server_api
//80001语音识别极速版(收费)
//80001 普通话 极速版输入法模型 有标点 支持自定义词库 http://vop.baidu.com/pro_api
var options = new Dictionary<string, object>
{
{"dev_pid", 1536}
};
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, type.TrimStart('.'), 16000, options);
comboBox9.Text = Convert.ToString(result);
result.TryGetValue("result", out JToken resultStr);
if (Convert.ToString(resultStr) != "")
{
comboBox9.Text = Convert.ToString(resultStr);
}
}
/// <summary>
/// 音频文件转语音
/// </summary>
/// <param name="filepath"></param>
/// <returns></returns>
public static string BaiduTranslateToText(string filepath, string language, string para_HZ)
{
string returnStr = "";
if (string.IsNullOrEmpty(filepath))
{
returnStr = "文件不存在";
}
else
{
string format = Path.GetExtension(filepath).Substring(1);
int rate = int.Parse(para_HZ);
string cuid = para_API_id;
FileStream fs = new FileStream(filepath, FileMode.Open);
byte[] voice = new byte[fs.Length];
fs.Read(voice, 0, voice.Length);
fs.Close();
Dictionary <string, object> options = new Dictionary <string, object>();
if (language == "zh")
{
options.Add("dev_pid", 1536);
}
else
{
options.Add("dev_pid", 1737);
}
Asr asr = new Asr(Api_Key, Secret_Key);
var result = asr.Recognize(voice, format, rate, options);
JObject obj = JObject.Parse(Newtonsoft.Json.JsonConvert.SerializeObject(result));
if (obj["err_msg"].Value <string>() == "success.")
{
returnStr = obj["result"][0].ToString();
}
else
{
returnStr = "3301-百度语音转文字出错";
}
}
return(returnStr);
}
public string Baidu_RecognizeByAPI(string filePhysicsPath)
{
new FFmpegSerializeOperator().ConvertToAmr(Parameters.CurrentDirectory, filePhysicsPath, filePhysicsPath);
Asr _asrClient = new Asr("TmZSp2x7nZXlPIW6GapK45XG", "BuWaLotBu4ho1GnblaMbue3ppryGuOYG ");
var data = File.ReadAllBytes(filePhysicsPath);
Dictionary <string, object> dic = new Dictionary <string, object>();
dic.Add("lan", "en");
var result = _asrClient.Recognize(data, "WAV", 8000);
string err_msg = result.GetValue("err_msg").ToString();
string content = string.Empty;
if (string.Equals(err_msg, "success."))
{
content = result.GetValue("result").ToString();
}
if (!string.Equals("success.", err_msg) || content.Split('"').Length >= 5)
{
result = Baidu_RecognizeByAPI(filePhysicsPath, true);
}
return(result.GetValue("result").ToString());
}
