Alibaba Cloud Model Studio:Real-time speech recognition: Fun-ASR/Paraformer

import com.alibaba.dashscope.audio.asr.recognition.Recognition;
import com.alibaba.dashscope.audio.asr.recognition.RecognitionParam;
import com.alibaba.dashscope.audio.asr.recognition.RecognitionResult;
import com.alibaba.dashscope.common.ResultCallback;
import com.alibaba.dashscope.utils.Constants;

import javax.sound.sampled.AudioFormat;
import javax.sound.sampled.AudioSystem;
import javax.sound.sampled.TargetDataLine;

import java.nio.ByteBuffer;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

public class Main {
    public static void main(String[] args) throws InterruptedException {
        // URL berikut untuk wilayah Singapura. Jika Anda menggunakan model di wilayah Beijing, ganti URL dengan wss://dashscope.aliyuncs.com/api-ws/v1/inference.
        Constants.baseWebsocketApiUrl = "wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference";
        ExecutorService executorService = Executors.newSingleThreadExecutor();
        executorService.submit(new RealtimeRecognitionTask());
        executorService.shutdown();
        executorService.awaitTermination(1, TimeUnit.MINUTES);
        System.exit(0);
    }
}

class RealtimeRecognitionTask implements Runnable {
    @Override
    public void run() {
        RecognitionParam param = RecognitionParam.builder()
                .model("fun-asr-realtime")
                // Kunci API untuk wilayah Singapura dan Beijing berbeda. Untuk mendapatkan Kunci API, lihat https://www.alibabacloud.com/help/en/model-studio/get-api-key.
                // Jika Anda belum mengonfigurasi variabel lingkungan, ganti baris berikut dengan Kunci API Anda: .apiKey("sk-xxx")
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .format("pcm")
                .sampleRate(16000)
                .build();
        Recognition recognizer = new Recognition();

        ResultCallback<RecognitionResult> callback = new ResultCallback<RecognitionResult>() {
            @Override
            public void onEvent(RecognitionResult result) {
                if (result.isSentenceEnd()) {
                    System.out.println("Final Result: " + result.getSentence().getText());
                } else {
                    System.out.println("Intermediate Result: " + result.getSentence().getText());
                }
            }

            @Override
            public void onComplete() {
                System.out.println("Recognition complete");
            }

            @Override
            public void onError(Exception e) {
                System.out.println("RecognitionCallback error: " + e.getMessage());
            }
        };
        try {
            recognizer.call(param, callback);
            // Buat AudioFormat.
            AudioFormat audioFormat = new AudioFormat(16000, 16, 1, true, false);
            // Dapatkan target data line yang sesuai dengan format yang ditentukan.
            TargetDataLine targetDataLine =
                    AudioSystem.getTargetDataLine(audioFormat);
            targetDataLine.open(audioFormat);
            // Mulai merekam.
            targetDataLine.start();
            ByteBuffer buffer = ByteBuffer.allocate(1024);
            long start = System.currentTimeMillis();
            // Rekam selama 50 detik dan lakukan transkripsi ucapan real-time.
            while (System.currentTimeMillis() - start < 50000) {
                int read = targetDataLine.read(buffer.array(), 0, buffer.capacity());
                if (read > 0) {
                    buffer.limit(read);
                    // Kirim data audio yang direkam ke layanan pengenalan streaming.
                    recognizer.sendAudioFrame(buffer);
                    buffer = ByteBuffer.allocate(1024);
                    // Tidur sejenak untuk mencegah penggunaan CPU tinggi.
                    Thread.sleep(20);
                }
            }
            recognizer.stop();
        } catch (Exception e) {
            e.printStackTrace();
        } finally {
            // Tutup koneksi WebSocket setelah tugas selesai.
            recognizer.getDuplexApi().close(1000, "bye");
        }

        System.out.println(
                "[Metric] requestId: "
                        + recognizer.getLastRequestId()
                        + ", first package delay ms: "
                        + recognizer.getFirstPackageDelay()
                        + ", last package delay ms: "
                        + recognizer.getLastPackageDelay());
    }
}

import os
import signal  # Untuk penanganan event keyboard (tekan "Ctrl+C" untuk menghentikan perekaman).
import sys

import dashscope
import pyaudio
from dashscope.audio.asr import *

mic = None
stream = None

# Atur parameter perekaman.
sample_rate = 16000  # Laju pengambilan sampel (Hz).
channels = 1  # Saluran mono.
dtype = 'int16'  # Tipe data.
format_pcm = 'pcm'  # Format data audio.
block_size = 3200  # Jumlah frame per buffer.


# Definisikan callback pengenalan ucapan real-time.
class Callback(RecognitionCallback):
    def on_open(self) -> None:
        global mic
        global stream
        print('RecognitionCallback open.')
        mic = pyaudio.PyAudio()
        stream = mic.open(format=pyaudio.paInt16,
                          channels=1,
                          rate=16000,
                          input=True)

    def on_close(self) -> None:
        global mic
        global stream
        print('RecognitionCallback close.')
        stream.stop_stream()
        stream.close()
        mic.terminate()
        stream = None
        mic = None

    def on_complete(self) -> None:
        print('RecognitionCallback completed.')  # Pengenalan selesai.

    def on_error(self, message) -> None:
        print('RecognitionCallback task_id: ', message.request_id)
        print('RecognitionCallback error: ', message.message)
        # Hentikan dan tutup aliran audio jika sedang berjalan.
        if 'stream' in globals() and stream.active:
            stream.stop()
            stream.close()
        # Keluar dari program secara paksa.
        sys.exit(1)

    def on_event(self, result: RecognitionResult) -> None:
        sentence = result.get_sentence()
        if 'text' in sentence:
            print('RecognitionCallback text: ', sentence['text'])
            if RecognitionResult.is_sentence_end(sentence):
                print(
                    'RecognitionCallback sentence end, request_id:%s, usage:%s'
                    % (result.get_request_id(), result.get_usage(sentence)))


def signal_handler(sig, frame):
    print('Ctrl+C ditekan, hentikan pengenalan ...')
    # Hentikan pengenalan.
    recognition.stop()
    print('Pengenalan dihentikan.')
    print(
        '[Metric] requestId: {}, first package delay ms: {}, last package delay ms: {}'
        .format(
            recognition.get_last_request_id(),
            recognition.get_first_package_delay(),
            recognition.get_last_package_delay(),
        ))
    # Keluar dari program secara paksa.
    sys.exit(0)


# Definisikan fungsi utama.
if __name__ == '__main__':
    # Kunci API untuk wilayah Singapura dan Beijing berbeda. Untuk mendapatkan Kunci API, lihat https://www.alibabacloud.com/help/en/model-studio/get-api-key
    # Jika Anda belum mengonfigurasi variabel lingkungan, ganti baris berikut dengan Kunci API Model Studio Anda: dashscope.api_key = "sk-xxx"
    dashscope.api_key = os.environ.get('DASHSCOPE_API_KEY')

    # Berikut adalah URL untuk wilayah Singapura. Jika Anda menggunakan model di wilayah Beijing, ganti URL dengan: wss://dashscope.aliyuncs.com/api-ws/v1/inference
    dashscope.base_websocket_api_url='wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference'

    # Buat callback pengenalan.
    callback = Callback()

    # Panggil layanan pengenalan dalam mode asinkron. Anda dapat menyesuaikan parameter pengenalan, seperti model, format, dan sample_rate.
    recognition = Recognition(
        model='fun-asr-realtime',
        format=format_pcm,
        # Format yang didukung meliputi 'pcm', 'wav', 'opus', 'speex', 'aac', dan 'amr'. Untuk informasi lebih lanjut, lihat dokumentasi.
        sample_rate=sample_rate,
        # Mendukung 8000 dan 16000.
        semantic_punctuation_enabled=False,
        callback=callback)

    # Mulai pengenalan.
    recognition.start()

    signal.signal(signal.SIGINT, signal_handler)
    print("Tekan 'Ctrl+C' untuk menghentikan perekaman dan pengenalan...")
    # Atur penangan sinyal untuk menangkap sinyal interupsi (Ctrl+C).

    while True:
        if stream:
            data = stream.read(3200, exception_on_overflow=False)
            recognition.send_audio_frame(data)
        else:
            break

    recognition.stop()

File audio yang digunakan dalam contoh adalah asr_example.wav.

import com.alibaba.dashscope.api.GeneralApi;
import com.alibaba.dashscope.audio.asr.recognition.Recognition;
import com.alibaba.dashscope.audio.asr.recognition.RecognitionParam;
import com.alibaba.dashscope.audio.asr.recognition.RecognitionResult;
import com.alibaba.dashscope.base.HalfDuplexParamBase;
import com.alibaba.dashscope.common.GeneralListParam;
import com.alibaba.dashscope.common.ResultCallback;
import com.alibaba.dashscope.protocol.GeneralServiceOption;
import com.alibaba.dashscope.protocol.HttpMethod;
import com.alibaba.dashscope.protocol.Protocol;
import com.alibaba.dashscope.protocol.StreamingMode;
import com.alibaba.dashscope.utils.Constants;

import java.io.FileInputStream;
import java.nio.ByteBuffer;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.time.LocalDateTime;
import java.time.format.DateTimeFormatter;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

class TimeUtils {
    private static final DateTimeFormatter formatter =
            DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss.SSS");

    public static String getTimestamp() {
        return LocalDateTime.now().format(formatter);
    }
}

public class Main {
    public static void main(String[] args) throws InterruptedException {
        // URL berikut untuk wilayah Singapura. Jika Anda menggunakan model di wilayah Beijing, ganti URL dengan wss://dashscope.aliyuncs.com/api-ws/v1/inference.
        Constants.baseWebsocketApiUrl = "wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference";
        // Di lingkungan produksi, Anda hanya perlu memanggil metode ini sekali saat startup aplikasi.
        warmUp();

        ExecutorService executorService = Executors.newSingleThreadExecutor();
        executorService.submit(new RealtimeRecognitionTask(Paths.get(System.getProperty("user.dir"), "asr_example.wav")));
        executorService.shutdown();

        // tunggu hingga semua tugas selesai
        executorService.awaitTermination(1, TimeUnit.MINUTES);
        System.exit(0);
    }

    public static void warmUp() {
        try {
            // Permintaan GET ringan untuk membuat koneksi
            GeneralServiceOption warmupOption = GeneralServiceOption.builder()
                    .protocol(Protocol.HTTP)
                    .httpMethod(HttpMethod.GET)
                    .streamingMode(StreamingMode.OUT)
                    .path("assistants")
                    .build();

            warmupOption.setBaseHttpUrl(Constants.baseHttpApiUrl);
            GeneralApi<HalfDuplexParamBase> api = new GeneralApi<>();
            api.get(GeneralListParam.builder().limit(1L).build(), warmupOption);
        } catch (Exception e) {
            // Setel ulang flag untuk memungkinkan percobaan ulang jika pemanasan awal gagal
        }
    }
}

class RealtimeRecognitionTask implements Runnable {
    private Path filepath;

    public RealtimeRecognitionTask(Path filepath) {
        this.filepath = filepath;
    }

    @Override
    public void run() {
        RecognitionParam param = RecognitionParam.builder()
                .model("fun-asr-realtime")
                // Kunci API untuk wilayah Singapura dan Beijing berbeda. Untuk mendapatkan Kunci API, lihat https://www.alibabacloud.com/help/en/model-studio/get-api-key.
                // Jika Anda belum mengonfigurasi variabel lingkungan, ganti baris berikut dengan Kunci API Anda: .apiKey("sk-xxx")
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .format("wav")
                .sampleRate(16000)
                .build();
        Recognition recognizer = new Recognition();

        String threadName = Thread.currentThread().getName();

        ResultCallback<RecognitionResult> callback = new ResultCallback<RecognitionResult>() {
            @Override
            public void onEvent(RecognitionResult message) {
                if (message.isSentenceEnd()) {

                    System.out.println(TimeUtils.getTimestamp()+" "+
                            "[process " + threadName + "] Final Result:" + message.getSentence().getText());
                } else {
                    System.out.println(TimeUtils.getTimestamp()+" "+
                            "[process " + threadName + "] Intermediate Result: " + message.getSentence().getText());
                }
            }

            @Override
            public void onComplete() {
                System.out.println(TimeUtils.getTimestamp()+" "+"[" + threadName + "] Recognition complete");
            }

            @Override
            public void onError(Exception e) {
                System.out.println(TimeUtils.getTimestamp()+" "+
                        "[" + threadName + "] RecognitionCallback error: " + e.getMessage());
            }
        };

        try {
            recognizer.call(param, callback);
            // Harap ganti path dengan path file audio Anda
            System.out.println(TimeUtils.getTimestamp()+" "+"[" + threadName + "] Input file_path is: " + this.filepath);
            // Baca file dan kirim audio per chunk
            FileInputStream fis = new FileInputStream(this.filepath.toFile());
            byte[] allData = new byte[fis.available()];
            int ret = fis.read(allData);
            fis.close();

            int sendFrameLength = 3200;
            for (int i = 0; i * sendFrameLength < allData.length; i ++) {
                int start = i * sendFrameLength;
                int end = Math.min(start + sendFrameLength, allData.length);
                ByteBuffer byteBuffer = ByteBuffer.wrap(allData, start, end - start);
                recognizer.sendAudioFrame(byteBuffer);
                Thread.sleep(100);
            }

            System.out.println(TimeUtils.getTimestamp()+" "+LocalDateTime.now());
            recognizer.stop();
        } catch (Exception e) {
            e.printStackTrace();
        } finally {
            // Tutup koneksi WebSocket setelah tugas selesai.
            recognizer.getDuplexApi().close(1000, "bye");
        }

        System.out.println(
                "["
                        + threadName
                        + "][Metric] requestId: "
                        + recognizer.getLastRequestId()
                        + ", first package delay ms: "
                        + recognizer.getFirstPackageDelay()
                        + ", last package delay ms: "
                        + recognizer.getLastPackageDelay());
    }
}

File audio yang digunakan dalam contoh adalah: asr_example.wav.

import os
import time
import dashscope
from dashscope.audio.asr import *

# Kunci API untuk wilayah Singapura dan Beijing berbeda. Untuk mendapatkan Kunci API, lihat https://www.alibabacloud.com/help/en/model-studio/get-api-key
# Jika Anda belum mengonfigurasi variabel lingkungan, ganti baris berikut dengan Kunci API Model Studio Anda: dashscope.api_key = "sk-xxx"
dashscope.api_key = os.environ.get('DASHSCOPE_API_KEY')

# Berikut adalah URL untuk wilayah Singapura. Jika Anda menggunakan model di wilayah Beijing, ganti URL dengan: wss://dashscope.aliyuncs.com/api-ws/v1/inference
dashscope.base_websocket_api_url='wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference'

from datetime import datetime


def get_timestamp():
    now = datetime.now()
    formatted_timestamp = now.strftime("[%Y-%m-%d %H:%M:%S.%f]")
    return formatted_timestamp


class Callback(RecognitionCallback):
    def on_complete(self) -> None:
        print(get_timestamp() + ' Recognition completed')  # Pengenalan selesai.

    def on_error(self, result: RecognitionResult) -> None:
        print('Recognition task_id: ', result.request_id)
        print('Recognition error: ', result.message)
        exit(0)

    def on_event(self, result: RecognitionResult) -> None:
        sentence = result.get_sentence()
        if 'text' in sentence:
            print(get_timestamp() + ' RecognitionCallback text: ', sentence['text'])
        if RecognitionResult.is_sentence_end(sentence):
            print(get_timestamp() +
                  'RecognitionCallback sentence end, request_id:%s, usage:%s'
                  % (result.get_request_id(), result.get_usage(sentence)))


callback = Callback()

recognition = Recognition(model='fun-asr-realtime',
                          format='wav',
                          sample_rate=16000,
                          callback=callback)

try:
    audio_data: bytes = None
    f = open("asr_example.wav", 'rb')
    if os.path.getsize("asr_example.wav"):
        # Baca seluruh file ke dalam buffer sekaligus.
        file_buffer = f.read()
        f.close()
        print("Start Recognition")
        recognition.start()

        # Kirim data dalam potongan 3200 byte dari buffer.
        buffer_size = len(file_buffer)
        offset = 0
        chunk_size = 3200

        while offset < buffer_size:
            # Hitung ukuran potongan data saat ini yang akan dikirim.
            remaining_bytes = buffer_size - offset
            current_chunk_size = min(chunk_size, remaining_bytes)

            # Ekstrak potongan data saat ini dari buffer.
            audio_data = file_buffer[offset:offset + current_chunk_size]

            # Kirim frame data audio.
            recognition.send_audio_frame(audio_data)
            # Perbarui offset.
            offset += current_chunk_size

            # Tambahkan jeda untuk mensimulasikan transmisi real-time.
            time.sleep(0.1)

        recognition.stop()
    else:
        raise Exception(
            'File yang diberikan kosong (panjang nol byte)')
except Exception as e:
    raise e

print(
    '[Metric] requestId: {}, first package delay ms: {}, last package delay ms: {}'
    .format(
        recognition.get_last_request_id(),
        recognition.get_first_package_delay(),
        recognition.get_last_package_delay(),
    ))