ai_qback/spoken_language/soeexample.py

# -*- coding: utf-8 -*-
import json
import os
import threading
import time

import requests

from spoken_language.common import credential
from spoken_language.read_config import read_config
from spoken_language.soe import speaking_assessment

config_data = read_config()
app_id, secret_id, secret_key = config_data['appId'], config_data['SecretId'], config_data['SecretKey']

APPID = app_id
SECRET_ID = secret_id
SECRET_KEY = secret_key
TOKEN = ""
ENGINE_MODEL_TYPE = "16k_en"
SLICE_SIZE = 32000

spoken_result = {}


class MySpeechRecognitionListener(speaking_assessment.SpeakingAssessmentListener):
    def __init__(self, id):
        self.id = id

    def on_recognition_start(self, response):
        pass

    def on_intermediate_result(self, response):
        rsp_str = json.dumps(response, ensure_ascii=False)

    def on_recognition_complete(self, response):
        global spoken_result
        spoken_result[self.id] = response

    def on_fail(self, response):
        rsp_str = json.dumps(response, ensure_ascii=False)


def process(id):
    audio = r"C:\Users\86131\Desktop\音频\output_16k_mono.mp3"
    listener = MySpeechRecognitionListener(id)

    credential_var = credential.Credential(SECRET_ID, SECRET_KEY)
    recognizer = speaking_assessment.SpeakingAssessment(
        APPID, credential_var, ENGINE_MODEL_TYPE, listener)
    recognizer.set_text_mode(0)
    recognizer.set_ref_text("anyway")
    recognizer.set_eval_mode(0)
    recognizer.set_keyword("")
    recognizer.set_sentence_info_enabled(0)
    recognizer.set_voice_format(1)
    try:
        recognizer.start()
        with open(audio, 'rb') as f:
            content = f.read(SLICE_SIZE)
            while content:
                recognizer.write(content)
                content = f.read(SLICE_SIZE)
                # sleep模拟实际实时语音发送间隔

                time.sleep(0.2)
    except Exception as e:
        print(e)
    finally:
        recognizer.stop()


def process_rec(task_id, audio_path, audio_text, audio_binary=None):
    audio = audio_path
    listener = MySpeechRecognitionListener(task_id)

    credential_var = credential.Credential(SECRET_ID, SECRET_KEY)
    recognizer = speaking_assessment.SpeakingAssessment(
        APPID, credential_var, ENGINE_MODEL_TYPE, listener)
    recognizer.set_text_mode(0)
    recognizer.set_ref_text(audio_text)
    recognizer.set_eval_mode(1)
    recognizer.set_keyword("")
    recognizer.set_sentence_info_enabled(0)
    recognizer.set_voice_format(2)

    recognizer.set_rec_mode(1)
    try:
        recognizer.start()
        if audio_binary:

            recognizer.write(audio_binary)
        else:
            with open(f"{task_id}.mp3", 'rb') as f:
                content = f.read()
                recognizer.write(content)
    except Exception as e:
        print(e)
    finally:
        recognizer.stop()


def process_multithread(number):
    thread_list = []
    for i in range(0, number):
        thread = threading.Thread(target=process, args=(i,))
        thread_list.append(thread)
        thread.start()

    for thread in thread_list:
        thread.join()


def make_spoken(task_id, audio_url, audio_content, audio_text):
    if audio_url:
        print("有url,应该去下载mp3文件")

        r = requests.get(audio_url)
        audio_content = r.content
    else:
        with open(f"{task_id}.mp3", 'wb') as f:
            f.write(audio_content)

    process_rec(task_id, audio_path=f"", audio_text=audio_text, audio_binary=audio_content)
    global spoken_result
    for _ in range(60):
        if task_id in spoken_result:
            r = spoken_result[task_id]
            del spoken_result[task_id]
            if os.path.exists(f"{task_id}.mp3"):
                os.remove(f"{task_id}.mp3")
            return r
        time.sleep(0.5)
    return None


if __name__ == "__main__":
    process_rec(0, r"C:\Users\86131\Desktop\音频\output_16k_mono.mp3", "You must study to be frank with the world apple")