Realitykit结合Speech实现语音控制AR机器人移动(完整代码)

利用Apple最新的Realitykit搭配ARkit实现虚拟物体的放置，结合内置的Speech库实现语音的识别功能，将语音内容转为文本内容，从而让机器进行运动。

大体思路：

1、配置并启动ARkit环境。

2、构建Entity实体。可以用Apple官方的CreatingAPhotogrammetryCommandLineApp的代码文档来生成.usdz文件，从而建造自己想要的实体。

3、放置实体到现实环境中。通过raycast发射射线，通过转化获得现实世界中的x，y，z的坐标，从而把实体放在现实世界中。

4、实现机器人的运动。通过传入文本信息，利用实体的transition属性来进行移动。

5、通过SFSpeechRecognizer获得语音识别允许。

6、创造音频节点，将输入语音的设备设置为麦克风，将音频结果初始化，放到SFSpeechAudioBufferRecognitionRequest里面。

7、进行音频结果的处理，转化为文本，放到机器人运动中。

8、机器人完成运动

没介绍的其他代码就是为实体添加属性，让机器人能够移动，放缩，旋转。

还有一些手势的识别：长按添加新的实体，滑动让实体消失。

1.  
   import UIKit
2.  
   import RealityKit
3.  
   import ARKit
4.  
   import Speech
5.  
   class ViewController: UIViewController {
6.  
   @IBOutlet var arView: ARView!
7.  
    
8.  
   var entity: Entity?
9.  
   var moveToLocation: Transform = Transform()
10.  
    let moveTime: Double = 5
11.  
    //语音识别
12.  
    let speechRecognizer: SFSpeechRecognizer? = SFSpeechRecognizer()
13.  
    let speechRequest = SFSpeechAudioBufferRecognitionRequest()
14.  
    var speechTask = SFSpeechRecognitionTask()
15.  
    // 音频实例化
16.  
    let audioEngine = AVAudioEngine() //设立音频节点，处理音频输入和输出
17.  
    let audioSession = AVAudioSession() //音频记录初始化
18.  
    override func viewDidLoad() {
19.  
    super.viewDidLoad()
20.  
    entity = try! Entity.loadModel(named: "toy_robot_vintage.usdz")
21.  
    entity?.generateCollisionShapes(recursive: true)
22.  
    //包裹起来 增加碰撞属性
23.  
    arView.installGestures([.rotation,.scale,.translation], for: entity! as! HasCollision)
24.  
    //创建session
25.  
    startARSession()
26.  
    //创建手势 将2d位置转换为3d位置 获取位置传递到下面的函数中
27.  
    arView.addGestureRecognizer(UITapGestureRecognizer(target: self, action: #selector(handleTapLocation)))
28.  
    arView.addGestureRecognizer(UISwipeGestureRecognizer(target: self, action: #selector(handleSwipeLocation)))
29.  
    arView.addGestureRecognizer(UILongPressGestureRecognizer(target: self, action: #selector(handleLongPressLocation)))
30.  
    startSpeechRecognition()
31.  
     
32.  
    }
33.  
    @objc func handleTapLocation(_ recognizer:UITapGestureRecognizer){
34.  
    let tapLocation = recognizer.location(in: arView)
35.  
    //发射粒子 转化为3d坐标
36.  
    let result = arView.raycast(from: tapLocation, allowing: .estimatedPlane, alignment: .horizontal)
37.  
    //得到x,y,z坐标
38.  
    if let firstResult = result.first{
39.  
    let worldPosition = simd_make_float3(firstResult.worldTransform.columns.3)
40.  
    placeModelInWorld(object: entity!, position: worldPosition)
41.  
    }
42.  
    }
43.  
    @objc func handleSwipeLocation(_ recognizer: UISwipeGestureRecognizer){
44.  
    let longPressLocation = recognizer.location(in: arView)
45.  
    if let entity = arView.entity(at: longPressLocation){
46.  
    entity.anchor?.removeFromParent()
47.  
    }
48.  
    }
49.  
    @objc func handleLongPressLocation(_ recognizer: UILongPressGestureRecognizer){
50.  
    let doubleTapLocation = recognizer.location(in: arView)
51.  
    let result = arView.raycast(from: doubleTapLocation, allowing: .estimatedPlane, alignment: .horizontal)
52.  
    //得到x,y,z坐标
53.  
    if let firstResult = result.first{
54.  
    let worldPosition = simd_make_float3(firstResult.worldTransform.columns.3)
55.  
    if (arView.entity(at: doubleTapLocation) == nil){
56.  
    let objectAnchor = AnchorEntity(world: worldPosition)
57.  
    let entity1 = try! Entity.loadModel(named: "toy_robot_vintageOne.usdz")
58.  
    entity1.generateCollisionShapes(recursive: true)
59.  
    arView.installGestures([.translation,.rotation,.scale], for: entity1)
60.  
    objectAnchor.addChild(entity1)
61.  
    arView.scene.addAnchor(objectAnchor)
62.  
     
63.  
    }
64.  
    }
65.  
    }
66.  
    func startARSession(){
67.  
    arView.automaticallyConfigureSession = true
68.  
    let configuration = ARWorldTrackingConfiguration()
69.  
    configuration.planeDetection = [.horizontal]
70.  
    configuration.environmentTexturing = .automatic
71.  
    // arView.debugOptions = .showAnchorGeometry
72.  
    arView.session.run(configuration)
73.  
    }
74.  
    func placeModelInWorld(object:Entity,position:SIMD3<Float>){
75.  
    let objectAnchor = AnchorEntity(world: position)
76.  
    objectAnchor.addChild(object)
77.  
    arView.scene.addAnchor(objectAnchor)
78.  
    }
79.  
    func rebotMove(direction: String){
80.  
    switch direction{
81.  
    case "forward":
82.  
    moveToLocation.translation = (entity!.transform.translation) simd_float3(x:0,y:0,z:20)
83.  
    entity!.move(to: moveToLocation, relativeTo: entity!, duration: moveTime)
84.  
    walkAnimation(movementDuration: moveTime)
85.  
    print("moveForward")
86.  
    case "back":
87.  
    moveToLocation.translation = (entity!.transform.translation) simd_float3(x:0,y:0,z:-20)
88.  
    entity!.move(to: moveToLocation, relativeTo: entity!, duration: moveTime)
89.  
    walkAnimation(movementDuration: moveTime)
90.  
    case "left":
91.  
    let rotateToAngle = simd_quatf(angle: GLKMathDegreesToRadians(90), axis: SIMD3(x:0,y:1,z:0))
92.  
    entity!.setOrientation(rotateToAngle, relativeTo: entity!)
93.  
    case "right":
94.  
    let rotateToAngle = simd_quatf(angle: GLKMathDegreesToRadians(-90), axis: SIMD3(x:0,y:1,z:0))
95.  
    entity!.setOrientation(rotateToAngle, relativeTo: entity!)
96.  
    default:
97.  
    print("没有移动指令")
98.  
    }
99.  
    }
100.  
     func walkAnimation(movementDuration: Double){
101.  
     if let rebotAnimation = entity!.availableAnimations.first{
102.  
     entity!.playAnimation(rebotAnimation.repeat(duration: movementDuration),transitionDuration: 0.5,startsPaused: false)
103.  
     print("Yes")
104.  
     }else{
105.  
     print("没有相关动画")
106.  
     }
107.  
     }
108.  
     func startSpeechRecognition(){
109.  
     //获得允许
110.  
     requestPermission()
111.  
     //记录
112.  
     startAudioRecoding()
113.  
     //识别
114.  
     speechRecognize()
115.  
     }
116.  
     func requestPermission(){
117.  
     SFSpeechRecognizer.requestAuthorization { (authorizationStatus) in
118.  
     if(authorizationStatus == .authorized){
119.  
     print("允许")
120.  
     }else if(authorizationStatus == .denied){
121.  
     print("拒绝")
122.  
     }else if(authorizationStatus == .notDetermined){
123.  
     print("等待您的决定")
124.  
     }else if(authorizationStatus == .restricted){
125.  
     print("无法启用")
126.  
     }
127.  
     }
128.  
     }
129.  
     func startAudioRecoding(){
130.  
     //创造输入节点
131.  
     let node = audioEngine.inputNode
132.  
     let recordingFormate = node.outputFormat(forBus: 0)
133.  
     node.installTap(onBus: 0, bufferSize: 1024, format: recordingFormate) { (buffer, _) in
134.  
     self.speechRequest.append(buffer)
135.  
     }
136.  
     //启动引擎
137.  
      
138.  
     do{
139.  
     //配置音频会话为从麦克风录制
140.  
     try audioSession.setCategory(.record,mode: .measurement,options: .duckOthers)
141.  
     try audioSession.setActive(true,options: .notifyOthersOnDeactivation)
142.  
     audioEngine.prepare()
143.  
     try audioEngine.start()
144.  
     }
145.  
     catch{
146.  
      
147.  
     }
148.  
      
149.  
     }
150.  
     func speechRecognize(){
151.  
     guard let speechRecognizer = SFSpeechRecognizer()else{
152.  
     print("语音识别不可用")
153.  
     return
154.  
     }
155.  
     if (speechRecognizer.isAvailable == false){
156.  
     print("无法正常工作")
157.  
     }
158.  
     var count = 0
159.  
     speechTask = speechRecognizer.recognitionTask(with: speechRequest, resultHandler: { (result, error) in
160.  
     count = 1
161.  
     if(count == 1){
162.  
     guard let result = result else {
163.  
     return
164.  
     }
165.  
     let recognizedText = result.bestTranscription.segments.last
166.  
     self.rebotMove(direction: recognizedText!.substring)
167.  
     print(recognizedText!.substring)
168.  
     }else if(count>=3){
169.  
     count = 0
170.  
     }
171.  
     })
172.  
     }
173.  
     }

注释已经放在代码中了，只需要导入自己的.usdz文件就可以运行成功了。

这篇好文章是转载于：学新通技术网