Add SwiftUI demo project (#118)

* Commit after creating the project

* Add sherpa-onnx related files

* copy and modify files from sherpa-ncnn

* add app icon

ios-swiftui/SherpaOnnx/SherpaOnnx/Assets.xcassets/AccentColor.colorset/Contents.json

@@ -0,0 +1,11 @@
+{
+  "colors" : [
+    {
+      "idiom" : "universal"
+    }
+  ],
+  "info" : {
+    "author" : "xcode",
+    "version" : 1
+  }
+}

ios-swiftui/SherpaOnnx/SherpaOnnx/Assets.xcassets/AppIcon.appiconset/Contents.json

@@ -0,0 +1,14 @@
+{
+  "images" : [
+    {
+      "filename" : "k2-1024x1024.png",
+      "idiom" : "universal",
+      "platform" : "ios",
+      "size" : "1024x1024"
+    }
+  ],
+  "info" : {
+    "author" : "xcode",
+    "version" : 1
+  }
+}

ios-swiftui/SherpaOnnx/SherpaOnnx/Assets.xcassets/Contents.json

@@ -0,0 +1,6 @@
+{
+  "info" : {
+    "author" : "xcode",
+    "version" : 1
+  }
+}

ios-swiftui/SherpaOnnx/SherpaOnnx/ContentView.swift

@@ -0,0 +1,46 @@
+//
+//  ContentView.swift
+//  SherpaOnnx
+//
+//  Created by fangjun on 2023/4/5.
+//
+
+import SwiftUI
+
+struct ContentView: View {
+    @StateObject var sherpaOnnxVM = SherpaOnnxViewModel()
+
+    var body: some View {
+        VStack {
+            Text("ASR with Next-gen Kaldi")
+                .font(.title)
+            if sherpaOnnxVM.status == .stop {
+                Text("See https://github.com/k2-fsa/sherpa-onnx")
+                Text("Press the Start button to run!")
+            }
+            ScrollView(.vertical, showsIndicators: true) {
+                HStack {
+                    Text(sherpaOnnxVM.subtitles)
+                    Spacer()
+                }
+            }
+            Spacer()
+            Button {
+                toggleRecorder()
+            } label: {
+                Text(sherpaOnnxVM.status == .stop ? "Start" : "Stop")
+            }
+        }
+        .padding()
+    }
+
+    private func toggleRecorder() {
+        sherpaOnnxVM.toggleRecorder()
+    }
+}
+
+struct ContentView_Previews: PreviewProvider {
+    static var previews: some View {
+        ContentView()
+    }
+}

ios-swiftui/SherpaOnnx/SherpaOnnx/Extension.swift

@@ -0,0 +1,20 @@
+//
+//  Extension.swift
+//  SherpaOnnx
+//
+//  Created by knight on 2023/4/5.
+//
+
+import AVFoundation
+
+extension AudioBuffer {
+    func array() -> [Float] {
+        return Array(UnsafeBufferPointer(self))
+    }
+}
+
+extension AVAudioPCMBuffer {
+    func array() -> [Float] {
+        return self.audioBufferList.pointee.mBuffers.array()
+    }
+}

ios-swiftui/SherpaOnnx/SherpaOnnx/Model.swift

@@ -0,0 +1,35 @@
+import Foundation
+
+func getResource(_ forResource: String, _ ofType: String) -> String {
+  let path = Bundle.main.path(forResource: forResource, ofType: ofType)
+  precondition(
+    path != nil,
+    "\(forResource).\(ofType) does not exist!\n" + "Remember to change \n"
+      + "  Build Phases -> Copy Bundle Resources\n" + "to add it!"
+  )
+  return path!
+}
+/// Please refer to
+/// https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
+/// to download pre-trained models
+
+/// sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20 (Bilingual, Chinese + English)
+/// https://k2-fsa.github.io/sherpa/onnx/pretrained_models/zipformer-transducer-models.html
+func getBilingualStreamZhEnZipformer20230220() -> SherpaOnnxOnlineTransducerModelConfig {
+  let encoder = getResource("encoder-epoch-99-avg-1", "onnx")
+  let decoder = getResource("decoder-epoch-99-avg-1", "onnx")
+  let joiner = getResource("joiner-epoch-99-avg-1", "onnx")
+  let tokens = getResource("tokens", "txt")
+
+  return sherpaOnnxOnlineTransducerModelConfig(
+    encoder: encoder,
+    decoder: decoder,
+    joiner: joiner,
+    tokens: tokens,
+    numThreads: 2
+  )
+}
+
+/// Please refer to
+/// https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
+/// to add more models if you need

ios-swiftui/SherpaOnnx/SherpaOnnx/Preview Content/Preview Assets.xcassets/Contents.json

@@ -0,0 +1,6 @@
+{
+  "info" : {
+    "author" : "xcode",
+    "version" : 1
+  }
+}

ios-swiftui/SherpaOnnx/SherpaOnnx/SherpaOnnxApp.swift

@@ -0,0 +1,17 @@
+//
+//  SherpaOnnxApp.swift
+//  SherpaOnnx
+//
+//  Created by fangjun on 2023/4/5.
+//
+
+import SwiftUI
+
+@main
+struct SherpaOnnxApp: App {
+    var body: some Scene {
+        WindowGroup {
+            ContentView()
+        }
+    }
+}

ios-swiftui/SherpaOnnx/SherpaOnnx/SherpaOnnxViewModel.swift

@@ -0,0 +1,182 @@
+//
+//  SherpaOnnxViewModel.swift
+//  SherpaOnnx
+//
+//  Created by knight on 2023/4/5.
+//
+
+import Foundation
+import AVFoundation
+
+enum Status {
+    case stop
+    case recording
+}
+
+class SherpaOnnxViewModel: ObservableObject {
+    @Published var status: Status = .stop
+    @Published var subtitles: String = ""
+    
+    var sentences: [String] = []
+    
+    var audioEngine: AVAudioEngine? = nil
+    var recognizer: SherpaOnnxRecognizer! = nil
+    
+    var lastSentence: String = ""
+    let maxSentence: Int = 20
+    
+    var results: String {
+        if sentences.isEmpty && lastSentence.isEmpty {
+            return ""
+        }
+        if sentences.isEmpty {
+            return "0: \(lastSentence.lowercased())"
+        }
+
+        let start = max(sentences.count - maxSentence, 0)
+        if lastSentence.isEmpty {
+            return sentences.enumerated().map { (index, s) in "\(index): \(s.lowercased())" }[start...]
+                .joined(separator: "\n")
+        } else {
+            return sentences.enumerated().map { (index, s) in "\(index): \(s.lowercased())" }[start...]
+                .joined(separator: "\n") + "\n\(sentences.count): \(lastSentence.lowercased())"
+        }
+    }
+    
+    func updateLabel() {
+        DispatchQueue.main.async {
+            self.subtitles = self.results
+        }
+    }
+    
+    init() {
+        initRecognizer()
+        initRecorder()
+    }
+    
+    private func initRecognizer() {
+        // Please select one model that is best suitable for you.
+        //
+        // You can also modify Model.swift to add new pre-trained models from
+        // https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
+        let modelConfig = getBilingualStreamZhEnZipformer20230220()
+
+        let featConfig = sherpaOnnxFeatureConfig(
+            sampleRate: 16000,
+            featureDim: 80)
+
+        var config = sherpaOnnxOnlineRecognizerConfig(
+            featConfig: featConfig,
+            modelConfig: modelConfig,
+            enableEndpoint: true,
+            rule1MinTrailingSilence: 2.4,
+            rule2MinTrailingSilence: 0.8,
+            rule3MinUtteranceLength: 30,
+            decodingMethod: "greedy_search",
+            maxActivePaths: 4
+        )
+        recognizer = SherpaOnnxRecognizer(config: &config)
+    }
+    
+    private func initRecorder() {
+        print("init recorder")
+        audioEngine = AVAudioEngine()
+        let inputNode = self.audioEngine?.inputNode
+        let bus = 0
+        let inputFormat = inputNode?.outputFormat(forBus: bus)
+        let outputFormat = AVAudioFormat(
+            commonFormat: .pcmFormatFloat32,
+            sampleRate: 16000, channels: 1,
+            interleaved: false)!
+
+        let converter = AVAudioConverter(from: inputFormat!, to: outputFormat)!
+
+        inputNode!.installTap(
+            onBus: bus,
+            bufferSize: 1024,
+            format: inputFormat
+        ) {
+            (buffer: AVAudioPCMBuffer, when: AVAudioTime) in
+            var newBufferAvailable = true
+
+            let inputCallback: AVAudioConverterInputBlock = {
+                inNumPackets, outStatus in
+                if newBufferAvailable {
+                    outStatus.pointee = .haveData
+                    newBufferAvailable = false
+
+                    return buffer
+                } else {
+                    outStatus.pointee = .noDataNow
+                    return nil
+                }
+            }
+
+            let convertedBuffer = AVAudioPCMBuffer(
+                pcmFormat: outputFormat,
+                frameCapacity:
+                    AVAudioFrameCount(outputFormat.sampleRate)
+                * buffer.frameLength
+                / AVAudioFrameCount(buffer.format.sampleRate))!
+
+            var error: NSError?
+            let _ = converter.convert(
+                to: convertedBuffer,
+                error: &error, withInputFrom: inputCallback)
+
+            // TODO(fangjun): Handle status != haveData
+
+            let array = convertedBuffer.array()
+            if !array.isEmpty {
+                self.recognizer.acceptWaveform(samples: array)
+                while (self.recognizer.isReady()){
+                    self.recognizer.decode()
+                }
+                let isEndpoint = self.recognizer.isEndpoint()
+                let text = self.recognizer.getResult().text
+
+                if !text.isEmpty && self.lastSentence != text {
+                    self.lastSentence = text
+                    self.updateLabel()
+                    print(text)
+                }
+
+                if isEndpoint{
+                    if !text.isEmpty {
+                        let tmp = self.lastSentence
+                        self.lastSentence = ""
+                        self.sentences.append(tmp)
+                    }
+                    self.recognizer.reset()
+                }
+            }
+        }
+    }
+    
+    public func toggleRecorder() {
+        if status == .stop {
+            startRecorder()
+            status = .recording
+        } else {
+            stopRecorder()
+            status = .stop
+        }
+    }
+
+    private func startRecorder() {
+        lastSentence = ""
+        sentences = []
+
+        do {
+            try self.audioEngine?.start()
+        } catch let error as NSError {
+            print("Got an error starting audioEngine: \(error.domain), \(error)")
+        }
+        print("started")
+    }
+
+    private func stopRecorder() {
+        audioEngine?.stop()
+        print("stopped")
+    }
+}