一、什么是AIUI
AIUI 是一套人机交互解决方案，整合了语音唤醒、语音识别、语义理解、内容平台、语音合成（比普通的语音合成多一个发音人）等能力。
新用户有20个免费的装机量，每天有500交互次数
二、创建AIUI
点击服务管理进入AIUI的控制台
四、AIUI设置
默认自带有语音识别的，如果需要AIUI回答你，则需要开启语义理解，如果需要有声音则需要开启语音合成，否则调用SDK的时候是不会返回语义理解和语音合成的数据的
可以在上面的位置直接修改发音人，返回的语音合成数据也会改变
五、导入项目
1.下载SDK
2.我们需要将aiui.dll导入的Unity的Plugins文件夹中
3.我们需要将AIUI文件夹放到StreamingAssets文件夹中，这里面保存的一些配置信息，比如APPID，语音合成等等，只需要在程序开始时读取配置文件进行了，就不需要像讯飞的语音合成，语音识别等模块一样需要登录和登出了。
4.这里有一些工具类，是对 aiui.dll 的扩展，使用的是官方的 c# 的 demo
5.初始化配置
void Start()
{
// 为每一个设备设置对应唯一的SN（最好使用设备硬件信息(mac地址，设备序列号等）生成）
// 以便正确统计装机量，避免刷机或者应用卸载重装导致装机量重复计数
AIUISetting.setSystemInfo(AIUIConstant.KEY_SERIAL_NUM, GetMac());
string cfg = File.ReadAllText(Application.streamingAssetsPath + "\\AIUI\\cfg\\aiui.cfg");
agent = IAIUIAgent.Create(cfg, onEvent);
IAIUIMessage msg_start = IAIUIMessage.Create(AIUIConstant.CMD_START, 0, 0, "", IBuffer.Zero);
agent.SendMessage(msg_start);
msg_start.Destroy();
msg_start = null;
}
6.发送消息，这里是将麦克风录入的信息转换为byte[]，然后发送的到SDK中
public IEnumerator SendMassage(byte[] data)
{
//唤醒
IAIUIMessage msg_wakeup = IAIUIMessage.Create(AIUIConstant.CMD_WAKEUP, 0, 0, "", IBuffer.Zero);
agent.SendMessage(msg_wakeup);
msg_wakeup.Destroy();
msg_wakeup = null;
yield return new WaitForSeconds(0.2f);
print("wakeup");
//开始录音
IAIUIMessage msg_start_r = IAIUIMessage.Create(AIUIConstant.CMD_START_RECORD, 0, 0,
"data_type=audio,interact_mode=oneshot", IBuffer.Zero);
agent.SendMessage(msg_start_r);
msg_start_r.Destroy();
msg_start_r = null;
//发送语音
IBuffer buf_1 = IBuffer.FromData(data, data.Length);
IAIUIMessage msg_write_audio = IAIUIMessage.Create(AIUIConstant.CMD_WRITE, 0, 0, "data_type=audio", buf_1);
agent.SendMessage(msg_write_audio);
msg_write_audio.Destroy();
msg_write_audio = null;
buf_1 = null;
yield return new WaitForSeconds(0.04f);
}
7.返回消息
private void onEvent(IAIUIEvent ev)
{
switch (ev.GetEventType())
{
case AIUIConstant.EVENT_STATE:
{
switch (ev.GetArg1())
{
case AIUIConstant.STATE_IDLE:
{
print("EVENT_STATE: IDLE");
}
break;
case AIUIConstant.STATE_READY:
{
print("EVENT_STATE: READY");
}
break;
case AIUIConstant.STATE_WORKING:
{
print("EVENT_STATE: WORKING");
}
break;
}
}
break;
case AIUIConstant.EVENT_WAKEUP:
{
Debug.LogFormat("EVENT_WAKEUP: {0}", ev.GetInfo());
}
break;
case AIUIConstant.EVENT_SLEEP:
{
Debug.LogFormat("EVENT_WAKEUP: arg1={0}", ev.GetArg1());
}
break;
case AIUIConstant.EVENT_VAD:
{
switch (ev.GetArg1())
{
case AIUIConstant.VAD_BOS:
{
print("EVENT_VAD: BOS");
}
break;
case AIUIConstant.VAD_EOS:
{
print("EVENT_VAD: EOS");
}
break;
}
}
break;
case AIUIConstant.EVENT_RESULT://返回成功时传入的数据
{
try
{
var info = JsonConvert.DeserializeObject(ev.GetInfo());
vAR Datas = info["data"] as JArray;
var data = datas[0] as JObject;
var param = data["params"] as JObject;
var contents = data["content"] as JArray;
var content = contents[0] as JObject;
string sub = param["sub"].ToString();
string cnt_id = content["cnt_id"].ToString();
int dataLen = 0;
byte[] buffer = ev.GetData().GetBinary(cnt_id, ref dataLen);
Debug.LogFormat("sub: {0} ，info: {1}", sub, ev.GetInfo());
switch (sub)
{
case "iat":
string jsonObject = Encoding.UTF8.GetString(buffer).Replace('\0', ' ');
JsonData deJson = JsonMapper.ToObject(jsonObject);
StringBuilder cont = new StringBuilder();
foreach (JsonData item in deJson["text"]["ws"])
{
cont.Append(item["cw"][0]["w"]);
}
//切换的主线程
Loom.QueueOnMainThread(() => {
sendText.text = cont.ToString();
});
break;
case "nlp":
string jsonObject1 = Encoding.UTF8.GetString(buffer).Replace('\0',' ');
JsonData deJson1 = JsonMapper.ToObject(jsonObject1);
//切换的主线程
Loom.QueueOnMainThread(() => {
returnText.text = deJson1["intent"]["answer"]["text"].ToString();
});
break;
case "tts":
string jsonObject2 = ev.GetInfo().Replace('\0', ' ');
JsonData deJson2 = JsonMapper.ToObject(jsonObject2);
//复制数据
for (int i = 0; i {
Save();
//audioData = new List();
//语音合成
//audioSource.clip = Clip(audioData.ToArray());
//audioSource.loop = false;
//audioSource.Play();
});
}
break;
case "asr":
break;
default:
break;
}
datas = null;
data = null;
param = null;
contents = null;
content = null;
info = null;
}
catch (Exception e)
{
print(e.Message); ;
}
}
break;
case AIUIConstant.EVENT_ERROR:
Debug.LogFormat("EVENT_ERROR: {0} {1}", ev.GetArg1(), ev.GetInfo());
break;
}
}
AIUIManager 全部代码
using aiui;
using LitJson;
using Newtonsoft.Json;
using Newtonsoft.Json.Linq;
using System;
using System.Collections;
using System.Collections.Generic;
using System.IO;
using System.Net.NetworkInformation;
using System.Runtime.InteropServices;
using System.Security.Cryptography;
using System.Text;
using UnityEngine;
using UnityEngine.UI;
//讯飞AIUI
public class AIUIManager : MonoBehaviour
{
public static AIUIManager Instance;
private IAIUIAgent agent;
public AudioSource audioSource;
public Text returnText;
public Text sendText;
private void Awake()
{
Instance = this;
}
// Start is called before the first frame update
void Start()
{
// 为每一个设备设置对应唯一的SN（最好使用设备硬件信息(mac地址，设备序列号等）生成）
// 以便正确统计装机量，避免刷机或者应用卸载重装导致装机量重复计数
AIUISetting.setSystemInfo(AIUIConstant.KEY_SERIAL_NUM, GetMac());
string cfg = File.ReadAllText(Application.streamingAssetsPath + "\\AIUI\\cfg\\aiui.cfg");
agent = IAIUIAgent.Create(cfg, onEvent);
IAIUIMessage msg_start = IAIUIMessage.Create(AIUIConstant.CMD_START, 0, 0, "", IBuffer.Zero);
agent.SendMessage(msg_start);
msg_start.Destroy();
msg_start = null;
}
public IEnumerator SendMassage(byte[] data)
{
//唤醒
IAIUIMessage msg_wakeup = IAIUIMessage.Create(AIUIConstant.CMD_WAKEUP, 0, 0, "", IBuffer.Zero);
agent.SendMessage(msg_wakeup);
msg_wakeup.Destroy();
msg_wakeup = null;
yield return new WaitForSeconds(0.2f);
print("wakeup");
//开始录音
IAIUIMessage msg_start_r = IAIUIMessage.Create(AIUIConstant.CMD_START_RECORD, 0, 0,
"data_type=audio,interact_mode=oneshot", IBuffer.Zero);
agent.SendMessage(msg_start_r);
msg_start_r.Destroy();
msg_start_r = null;
//发送语音
IBuffer buf_1 = IBuffer.FromData(data, data.Length);
IAIUIMessage msg_write_audio = IAIUIMessage.Create(AIUIConstant.CMD_WRITE, 0, 0, "data_type=audio", buf_1);
agent.SendMessage(msg_write_audio);
msg_write_audio.Destroy();
msg_write_audio = null;
buf_1 = null;
yield return new WaitForSeconds(0.04f);
}
public string GetMac()
{
NetworkInterface[] interfaces = NetworkInterface.GetAllNetworkInterfaces();
string mac = "";
foreach (NetworkInterface ni in interfaces)
{
if (ni.NetworkInterfaceType != NetworkInterfaceType.Loopback)
{
mac += ni.GetPhysicalAddress().ToString();
}
}
byte[] result = Encoding.Default.GetBytes(mac);
result = new MD5CryptoServiceProvider().ComputeHash(result);
StringBuilder builder = new StringBuilder();
for (int i = 0; i {
sendText.text = cont.ToString();
});
break;
case "nlp":
string jsonObject1 = Encoding.UTF8.GetString(buffer).Replace('\0',' ');
JsonData deJson1 = JsonMapper.ToObject(jsonObject1);
//切换的主线程
Loom.QueueOnMainThread(() => {
returnText.text = deJson1["intent"]["answer"]["text"].ToString();
});
break;
case "tts":
string jsonObject2 = ev.GetInfo().Replace('\0', ' ');
JsonData deJson2 = JsonMapper.ToObject(jsonObject2);
if (int.Parse(deJson2["data"][0]["content"][0]["dts"].ToString()) == 2)
{
for (int i = 0; i {
//Save();
//语音合成
audioSource.clip = Clip(audioData.ToArray());
//audioSource.loop = false;
audioSource.Play();
audioData = new List();
});
}
break;
case "asr":
break;
default:
break;
}
datas = null;
data = null;
param = null;
contents = null;
content = null;
info = null;
}
catch (Exception e)
{
print(e.Message); ;
}
}
break;
case AIUIConstant.EVENT_ERROR:
Debug.LogFormat("EVENT_ERROR: {0} {1}", ev.GetArg1(), ev.GetInfo());
break;
}
}
private List audioData = new List();
///
/// 把结构体转化为字节序列
///
/// 被转化的结构体
/// 返回字节序列
public static byte[] StructToBytes(object structure)
{
int size = Marshal.SizeOf(structure);
IntPtr buffer = Marshal.AllocHGlobal(size);
try
{
Marshal.StructureToPtr(structure, buffer, false);
Byte[] bytes = new Byte[size];
Marshal.Copy(buffer, bytes, 0, size);
return bytes;
}
finally
{
Marshal.FreeHGlobal(buffer);
}
}
///
/// 根据数据段的长度，生产文件头
///
/// 音频数据长度
/// 返回wav文件头结构体
public static WAVE_Header getWave_Header(int data_len)
{
WAVE_Header wav_Header = new WAVE_Header();
wav_Header.RIFF_ID = 0x46464952; //字符RIFF
wav_Header.File_Size = data_len + 36;
wav_Header.RIFF_Type = 0x45564157; //字符WAVE
wav_Header.FMT_ID = 0x20746D66; //字符fmt
wav_Header.FMT_Size = 16;
wav_Header.FMT_Tag = 0x0001;
wav_Header.FMT_Channel = 1; //单声道
wav_Header.FMT_SamplesPerSec = 16000; //采样频率
wav_Header.AvgBytesPerSec = 32000; //每秒所需字节数
wav_Header.BlockAlign = 2; //每个采样1个字节
wav_Header.BitsPerSample = 16; //每个采样8bit
wav_Header.DATA_ID = 0x61746164; //字符data
wav_Header.DATA_Size = data_len;
return wav_Header;
}
///
/// wave文件头
///
public struct WAVE_Header
{
public int RIFF_ID; //4 byte , 'RIFF'
public int File_Size; //4 byte , 文件长度
public int RIFF_Type; //4 byte , 'WAVE'
public int FMT_ID; //4 byte , 'fmt'
public int FMT_Size; //4 byte , 数值为16或18，18则最后又附加信息
public short FMT_Tag; //2 byte , 编码方式，一般为0x0001
public ushort FMT_Channel; //2 byte , 声道数目，1--单声道；2--双声道
public int FMT_SamplesPerSec; //4 byte , 采样频率
public int AvgBytesPerSec; //4 byte , 每秒所需字节数,记录每秒的数据量
public ushort BlockAlign; //2 byte , 数据块对齐单位(每个采样需要的字节数)
public ushort BitsPerSample; //2 byte , 每个采样需要的bit数
public int DATA_ID; //4 byte , 'data'
public int DATA_Size; //4 byte ,
}
#region 播放录音
public static AudioClip Clip(byte[] fileBytes, int offsetSamples = 0, string name = "ifly")
{
//string riff = Encoding.ASCII.GetString (fileBytes, 0, 4);
//string wave = Encoding.ASCII.GetString (fileBytes, 8, 4);
int subchunk1 = BitConverter.ToInt32(fileBytes, 16);
ushort audioFormat = BitConverter.ToUInt16(fileBytes, 20);
// NB: Only uncompressed PCM wav files are supported.
string formatCode = FormatCode(audioFormat);
//Debug.AssertFormat(audioFormat == 1 || audioFormat == 65534, "Detected format code '{0}' {1}, but only PCM and WaveFormatExtensable uncompressed formats are currently supported.", audioFormat, formatCode);
ushort channels = BitConverter.ToUInt16(fileBytes, 22);
int sampleRate = BitConverter.ToInt32(fileBytes, 24);
//int byteRate = BitConverter.ToInt32 (fileBytes, 28);
//UInt16 blockAlign = BitConverter.ToUInt16 (fileBytes, 32);
ushort bitDepth = BitConverter.ToUInt16(fileBytes, 34);
int headerOffset = 16 + 4 + subchunk1 + 4;
int subchunk2 = BitConverter.ToInt32(fileBytes, headerOffset);
//Debug.LogFormat ("riff={0} wave={1} subchunk1={2} format={3} channels={4} sampleRate={5} byteRate={6} blockAlign={7} bitDepth={8} headerOffset={9} subchunk2={10} filesize={11}", riff, wave, subchunk1, formatCode, channels, sampleRate, byteRate, blockAlign, bitDepth, headerOffset, subchunk2, fileBytes.Length);
//Log.Info(bitDepth);
float[] data;
switch (bitDepth)
{
case 8:
data = Convert8BitByteArrayToAudioClipData(fileBytes, headerOffset, subchunk2);
break;
case 16:
data = Convert16BitByteArrayToAudioClipData(fileBytes, headerOffset, subchunk2);
break;
case 24:
data = Convert24BitByteArrayToAudioClipData(fileBytes, headerOffset, subchunk2);
break;
case 32:
data = Convert32BitByteArrayToAudioClipData(fileBytes, headerOffset, subchunk2);
break;
default:
throw new Exception(bitDepth + " bit depth is not supported.");
}
AudioClip audioClip = AudioClip.Create(name, data.Length, channels, sampleRate, false);
audioClip.SetData(data, 0);
return audioClip;
}
private static string FormatCode(UInt16 code)
{
switch (code)
{
case 1:
return "PCM";
case 2:
return "ADPCM";
case 3:
return "IEEE";
case 7:
return "μ-law";
case 65534:
return "WaveFormatExtensable";
default:
Debug.LogWarning("Unknown wav code format:" + code);
return "";
}
}
#region wav file bytes to Unity AudioClip conversion methods
private static float[] Convert8BitByteArrayToAudioClipData(byte[] source, int headerOffset, int dataSize)
{
int wavSize = BitConverter.ToInt32(source, headerOffset);
headerOffset += sizeof(int);
Debug.AssertFormat(wavSize > 0 && wavSize == dataSize, "Failed to get valid 8-bit wav size: {0} from data bytes: {1} at offset: {2}", wavSize, dataSize, headerOffset);
float[] data = new float[wavSize];
sbyte maxValue = sbyte.MaxValue;
int i = 0;
while (i 0 && wavSize == dataSize, "Failed to get valid 16-bit wav size: {0} from data bytes: {1} at offset: {2}", wavSize, dataSize, headerOffset);
int x = sizeof(Int16); // block size = 2
int convertedSize = wavSize / x;
float[] data = new float[convertedSize];
Int16 maxValue = Int16.MaxValue;
int offset = 0;
int i = 0;
while (i 0 && wavSize == dataSize, "Failed to get valid 24-bit wav size: {0} from data bytes: {1} at offset: {2}", wavSize, dataSize, headerOffset);
int x = 3; // block size = 3
int convertedSize = wavSize / x;
int maxValue = Int32.MaxValue;
float[] data = new float[convertedSize];
byte[] block = new byte[sizeof(int)]; // using a 4 byte block for copying 3 bytes, then copy bytes with 1 offset
int offset = 0;
int i = 0;
while (i 0 && wavSize == dataSize, "Failed to get valid 32-bit wav size: {0} from data bytes: {1} at offset: {2}", wavSize, dataSize, headerOffset);
int x = sizeof(float); // block size = 4
int convertedSize = wavSize / x;
Int32 maxValue = Int32.MaxValue;
float[] data = new float[convertedSize];
int offset = 0;
int i = 0;
while (i
这里接收到消息时，是使用的其他线程，所以需要转换到主线程才能使用Unity的API,不然就会报错，我这里可以将传入的数据放在本地，也在内存中使用并播放（这里传输过来的数据是pcm，如果要在Unity中直接使用，需要写入WAV头才行）。语音合成的数据是多段传输的，所以需要整合在一起。
六、项目位置
链接：https://pan.baidu.com/s/1X83tzFAaRLFvPeleT2Pm0A
提取码：a1oj
只需要将SDK中的AIUI文件放在SteamingAssets文件夹中和替换aiui.dll文件即可。
