241205_给自己的应用加上语音助手功能

241205_给自己的应用加上语音助手功能

前面我们自己做了一个网易云音乐，但每次都要去点点点显得有点麻烦，所以我就考虑添加一些语音助手的功能。

其实当前在日常windows使用中，我觉得也就音乐播放需要一个语音助手交互，其他的功能，要么太复杂，简单的声控无法实现，要么语音控制的效率太低，不如直接上手操作。

最简单的语音助手仅需要实现三个功能，本地语音转文本，根据文本内容提取关键词完成对应指令，播放“收到”“已完成”等语音，我们此处就实现这样的最简单的语音助手。

语音转文字

本地语音转文字我们使用vosk语音识别库实现。model使用本地的一个60m的小模型.

这里附模型链接： https://pan.baidu.com/s/1mY2VoRqjSwS6n4qj1pPZMA?pwd=8888

解压放在项目根目录就可以了

为了实现一个较好的唤醒效果，我们采用后台实时录音的方式。检测到声音的时候就进行转文字，但是因为环境噪音是一直存在的，此时我们就要给他设定一个阈值，当超过这个阈值时，我们才认为他是有效输入，才进行转文字输入。

def SaveWave(model):
    # 设置音频参数
    FORMAT = pyaudio.paInt16    # 音频流的格式
    RATE = 44100                # 采样率，单位Hz
    CHUNK = 4000                # 单位帧
    THRESHOLDNUM = 10           # 静默时间，超过这个个数就保存文件
    THRESHOLD = 100             # 设定停止采集阈值

    audio = pyaudio.PyAudio()
    stream = audio.open(format=FORMAT,
                        channels=1,
                        rate=RATE,
                        input=True,
                        frames_per_buffer=CHUNK)
    frames = []
    print("开始录音...")
    count = 0
    while count < THRESHOLDNUM:
        data = stream.read(CHUNK, exception_on_overflow=False)
        np_data = np.frombuffer(data, dtype=np.int16)
        frame_energy = np.mean(np.abs(np_data))
        # print(frame_energy)
        # 如果能量低于阈值持续时间过长，则停止录音
        if frame_energy < THRESHOLD:
            count += 1
        elif count > 0:
            count -= 1

        frames.append(data)
    print("停止录音!")
    stream.stop_stream()
    stream.close()
    audio.terminate()

    rec = KaldiRecognizer(model, RATE)
    rec.SetWords(True)
    str_ret = ""
    for data in frames:
        if rec.AcceptWaveform(data):
            result = json.loads(rec.Result())
            if 'text' in result:
                str_ret += result['text']

    result = json.loads(rec.FinalResult())
    if 'text' in result:
        str_ret += result['text']

    str_ret = "".join(str_ret.split())
    return str_ret

执行命令

得到语音转文字的结果之后，我们就可指定对应执行的指令了，因为我们不需要联网去搜索结果，只需要定义我们自己能用到的指令就可以了，此处简单定义指令。

def weekup(mw):
    model = Model("vosk-model-cn-0.15")
    SetLogLevel(-1)

    while 1:
        res = SaveWave(model)
        if res != "" and res != None:
            print(res)
        if "小爱同学" in res:
            answer = "我在"
            answermethod(answer)

        if "聊天" in res:
            answer = "好的，请开始"
            answermethod(answer)
            from stt import run as stt_run
            result=stt_run()
        if "打开" in res:
            app_name=res.replace("打开",'').strip()
            if app_name in dir_path:
                app_path=dir_path[app_name]
                import subprocess
                subprocess.Popen(app_path)
                answer = "好的，已经为您打开了"
                answermethod(answer)
        if "歌" in res:
            from testyuncloud import yuncloudMW
            mw.playMusic()
            # print("播放歌曲")
        if "暂停" in res:
            mw.playMusic()
        if "上一首" in res:
            mw.prevMusic()
        if "下一首" in res:
            mw.nextMusic()

在这个方法中，我调用了一个answermethod方法，在这个方法中，我调用了百度的语音合成api，当然这里合成的都是一些简单的语音，你也可以直接自己录音，然后直接播放录音文件。就避免了调用api。

下面是answermethod以及百度语音合成api的调用。

def answermethod(answer):
    baidu_tts_test(answer)
    time.sleep(1)

# coding=utf-8
import sys
import json
import pygame
import time
import io
IS_PY3 = sys.version_info.major == 3
if IS_PY3:
    from urllib.request import urlopen
    from urllib.request import Request
    from urllib.error import URLError
    from urllib.parse import urlencode
    from urllib.parse import quote_plus
else:
    import urllib2
    from urllib import quote_plus
    from urllib2 import urlopen
    from urllib2 import Request
    from urllib2 import URLError
    from urllib import urlencode

API_KEY = ' ' # 这里和下面替换成你的百度语音合成服务api的key
SECRET_KEY = ' '

# TEXT = "欢迎使用百度语音合成。"

# 发音人选择, 基础音库：0为度小美，1为度小宇，3为度逍遥，4为度丫丫，
# 精品音库：5为度小娇，103为度米朵，106为度博文，110为度小童，111为度小萌，默认为度小美 
PER = 4
# 语速，取值0-15，默认为5中语速
SPD = 5
# 音调，取值0-15，默认为5中语调
PIT = 5
# 音量，取值0-9，默认为5中音量
VOL = 5
# 下载的文件格式, 3：mp3(default) 4： pcm-16k 5： pcm-8k 6. wav
AUE = 3

FORMATS = {3: "mp3", 4: "pcm", 5: "pcm", 6: "wav"}
FORMAT = FORMATS[AUE]

CUID = "123456PYTHON"

TTS_URL = 'http://tsn.baidu.com/text2audio'


class DemoError(Exception):
    pass


"""  TOKEN start """

TOKEN_URL = 'http://aip.baidubce.com/oauth/2.0/token'
SCOPE = 'audio_tts_post'  # 有此scope表示有tts能力，没有请在网页里勾选


def fetch_token():
    print("fetch token begin")
    params = {'grant_type': 'client_credentials',
              'client_id': API_KEY,
              'client_secret': SECRET_KEY}
    post_data = urlencode(params)
    if (IS_PY3):
        post_data = post_data.encode('utf-8')
    req = Request(TOKEN_URL, post_data)
    try:
        f = urlopen(req, timeout=5)
        result_str = f.read()
    except URLError as err:
        print('token http response http code : ' + str(err.code))
        result_str = err.read()
    if (IS_PY3):
        result_str = result_str.decode()

    print(result_str)
    result = json.loads(result_str)
    print(result)
    if ('access_token' in result.keys() and 'scope' in result.keys()):
        if not SCOPE in result['scope'].split(' '):
            raise DemoError('scope is not correct')
        print('SUCCESS WITH TOKEN: %s ; EXPIRES IN SECONDS: %s' % (result['access_token'], result['expires_in']))
        return result['access_token']
    else:
        raise DemoError('MAYBE API_KEY or SECRET_KEY not correct: access_token or scope not found in token response')


"""  TOKEN end """

def baidu_tts_test(text,PER=0,SPD=10,PIT=PIT,VOL=VOL):
    token = fetch_token()
    tex = quote_plus(text)  # 此处TEXT需要两次urlencode
    print(tex)
    params = {'tok': token, 'tex': tex, 'per': PER, 'spd': SPD, 'pit': PIT, 'vol': VOL, 'aue': AUE, 'cuid': CUID,
              'lan': 'zh', 'ctp': 1}  # lan ctp 固定参数

    data = urlencode(params)
    print('test on Web Browser' + TTS_URL + '?' + data)

    req = Request(TTS_URL, data.encode('utf-8'))
    has_error = False
    try:
        f = urlopen(req)
        result_str = f.read()

        headers = dict((name.lower(), value) for name, value in f.headers.items())

        has_error = ('content-type' not in headers.keys() or headers['content-type'].find('audio/') < 0)
    except  URLError as err:
        print('asr http response http code : ' + str(err.code))
        result_str = err.read()
        has_error = True

    save_file = "error.txt" if has_error else 'result.' + FORMAT
    with open(save_file, 'wb') as of:
        of.write(result_str)

    if has_error:
        if (IS_PY3):
            result_str = str(result_str, 'utf-8')
        print("tts api  error:" + result_str)
    else:
        # 初始化pygame
        pygame.mixer.init()

        # 将音频数据加载到内存
        pygame.mixer.music.load(io.BytesIO(result_str))

        # 播放音频
        pygame.mixer.music.play()

        # # 防止程序立即退出
        # while pygame.mixer.music.get_busy():
        #     time.sleep(1)
    print("result saved as :" + save_file)

随后我要实现调整音量的功能，但是识别出来调整音量的百分比是汉字，无法直接去调整，此处就要建立汉字的数字到阿拉伯数字的映射。

from pypinyin import lazy_pinyin

def hanzi_to_number_pinyin(hanzi_str):
    hnd = {
        'ling': 0, 'yi': 1, 'er': 2, 'san': 3, 'si': 4, 'wu': 5, 'liu': 6, 'qi': 7, 'ba': 8, 'jiu': 9,
        'shi': 10, 'bai': 100, 'qian': 1000, 'wan': 10000, 'yi': 100000000
    }
    pinyin_list = lazy_pinyin(hanzi_str)
    result = 0
    temp = 0
    for pinyin in pinyin_list:
        if pinyin in hnd:
            digit = hnd[pinyin]
            if digit >= 10:
                if temp == 0:
                    temp = 1
                result += temp * digit
                temp = 0
            else:
                temp = temp * 10 + digit
    result += temp
    return result

然后在关键词匹配的方法中实现调整音量

        if "音量" in res:
            volumn=res.split('百分之')[1]
            volumn=hanzi_to_number_pinyin(volumn)
            print(volumn)
            mw.volumeSet(volumn)

然后在我们的主方法里面，开启一个新线程，进行调用即可

以下是全部代码：

# weekupm.py
import json
import time

import pyaudio
import numpy as np
from vosk import Model, KaldiRecognizer, SetLogLevel
from baidu_tts import baidu_tts_test

dir_path={
        '浏览器': "C:\\Program Files (x86)\\Microsoft\\Edge\\Application\\msedge.exe",
        '记事本': "C:\\Windows\\System32\\notepad.exe",
        '计算器': "C:\\Windows\\System32\\calc.exe"
    }
def SaveWave(model):
    # 设置音频参数
    FORMAT = pyaudio.paInt16    # 音频流的格式
    RATE = 44100                # 采样率，单位Hz
    CHUNK = 4000                # 单位帧
    THRESHOLDNUM = 10           # 静默时间，超过这个个数就保存文件
    THRESHOLD = 100             # 设定停止采集阈值

    audio = pyaudio.PyAudio()
    stream = audio.open(format=FORMAT,
                        channels=1,
                        rate=RATE,
                        input=True,
                        frames_per_buffer=CHUNK)
    frames = []
    print("开始录音...")
    count = 0
    while count < THRESHOLDNUM:
        data = stream.read(CHUNK, exception_on_overflow=False)
        np_data = np.frombuffer(data, dtype=np.int16)
        frame_energy = np.mean(np.abs(np_data))
        # print(frame_energy)
        # 如果能量低于阈值持续时间过长，则停止录音
        if frame_energy < THRESHOLD:
            count += 1
        elif count > 0:
            count -= 1

        frames.append(data)
    print("停止录音!")
    stream.stop_stream()
    stream.close()
    audio.terminate()

    rec = KaldiRecognizer(model, RATE)
    rec.SetWords(True)
    str_ret = ""
    for data in frames:
        if rec.AcceptWaveform(data):
            result = json.loads(rec.Result())
            if 'text' in result:
                str_ret += result['text']

    result = json.loads(rec.FinalResult())
    if 'text' in result:
        str_ret += result['text']

    str_ret = "".join(str_ret.split())
    return str_ret

def answermethod(answer):
    baidu_tts_test(answer)
    time.sleep(1)
def weekup(mw):
    model = Model("vosk-model-cn-0.15")
    SetLogLevel(-1)

    while 1:
        res = SaveWave(model)
        if res != "" and res != None:
            print(res)
        if "小爱同学" in res:
            answer = "我在"
            answermethod(answer)

        if "聊天" in res:
            answer = "好的，请开始"
            answermethod(answer)
            from stt import run as stt_run
            result=stt_run()
        if "打开" in res:
            app_name=res.replace("打开",'').strip()
            if app_name in dir_path:
                app_path=dir_path[app_name]
                import subprocess
                subprocess.Popen(app_path)
                answer = "好的，已经为您打开了"
                answermethod(answer)
        if "歌" in res:
            from testyuncloud import yuncloudMW
            mw.playMusic()
            # print("播放歌曲")
        if "暂停" in res:
            mw.playMusic()
        if "上一首" in res:
            mw.prevMusic()
        if "下一首" in res:
            mw.nextMusic()
        if "音量" in res:
            volumn=res.split('百分之')[1]
            volumn=hanzi_to_number_pinyin(volumn)
            print(volumn)
            mw.volumeSet(volumn)


from pypinyin import lazy_pinyin

def hanzi_to_number_pinyin(hanzi_str):
    hnd = {
        'ling': 0, 'yi': 1, 'er': 2, 'san': 3, 'si': 4, 'wu': 5, 'liu': 6, 'qi': 7, 'ba': 8, 'jiu': 9,
        'shi': 10, 'bai': 100, 'qian': 1000, 'wan': 10000, 'yi': 100000000
    }
    pinyin_list = lazy_pinyin(hanzi_str)
    result = 0
    temp = 0
    for pinyin in pinyin_list:
        if pinyin in hnd:
            digit = hnd[pinyin]
            if digit >= 10:
                if temp == 0:
                    temp = 1
                result += temp * digit
                temp = 0
            else:
                temp = temp * 10 + digit
    result += temp
    return result

# baidu_tts.py
# coding=utf-8
import sys
import json
import pygame
import time
import io
IS_PY3 = sys.version_info.major == 3
if IS_PY3:
    from urllib.request import urlopen
    from urllib.request import Request
    from urllib.error import URLError
    from urllib.parse import urlencode
    from urllib.parse import quote_plus
else:
    import urllib2
    from urllib import quote_plus
    from urllib2 import urlopen
    from urllib2 import Request
    from urllib2 import URLError
    from urllib import urlencode

API_KEY = '2I46jXf5CmlNycyOVAETigud'
SECRET_KEY = 'IJu44Y0K09y93ZQAOsObkJEhOEd1NBQa'

# TEXT = "欢迎使用百度语音合成。"

# 发音人选择, 基础音库：0为度小美，1为度小宇，3为度逍遥，4为度丫丫，
# 精品音库：5为度小娇，103为度米朵，106为度博文，110为度小童，111为度小萌，默认为度小美 
PER = 4
# 语速，取值0-15，默认为5中语速
SPD = 5
# 音调，取值0-15，默认为5中语调
PIT = 5
# 音量，取值0-9，默认为5中音量
VOL = 5
# 下载的文件格式, 3：mp3(default) 4： pcm-16k 5： pcm-8k 6. wav
AUE = 3

FORMATS = {3: "mp3", 4: "pcm", 5: "pcm", 6: "wav"}
FORMAT = FORMATS[AUE]

CUID = "123456PYTHON"

TTS_URL = 'http://tsn.baidu.com/text2audio'


class DemoError(Exception):
    pass


"""  TOKEN start """

TOKEN_URL = 'http://aip.baidubce.com/oauth/2.0/token'
SCOPE = 'audio_tts_post'  # 有此scope表示有tts能力，没有请在网页里勾选


def fetch_token():
    print("fetch token begin")
    params = {'grant_type': 'client_credentials',
              'client_id': API_KEY,
              'client_secret': SECRET_KEY}
    post_data = urlencode(params)
    if (IS_PY3):
        post_data = post_data.encode('utf-8')
    req = Request(TOKEN_URL, post_data)
    try:
        f = urlopen(req, timeout=5)
        result_str = f.read()
    except URLError as err:
        print('token http response http code : ' + str(err.code))
        result_str = err.read()
    if (IS_PY3):
        result_str = result_str.decode()

    print(result_str)
    result = json.loads(result_str)
    print(result)
    if ('access_token' in result.keys() and 'scope' in result.keys()):
        if not SCOPE in result['scope'].split(' '):
            raise DemoError('scope is not correct')
        print('SUCCESS WITH TOKEN: %s ; EXPIRES IN SECONDS: %s' % (result['access_token'], result['expires_in']))
        return result['access_token']
    else:
        raise DemoError('MAYBE API_KEY or SECRET_KEY not correct: access_token or scope not found in token response')


"""  TOKEN end """

def baidu_tts_test(text,PER=0,SPD=10,PIT=PIT,VOL=VOL):
    token = fetch_token()
    tex = quote_plus(text)  # 此处TEXT需要两次urlencode
    print(tex)
    params = {'tok': token, 'tex': tex, 'per': PER, 'spd': SPD, 'pit': PIT, 'vol': VOL, 'aue': AUE, 'cuid': CUID,
              'lan': 'zh', 'ctp': 1}  # lan ctp 固定参数

    data = urlencode(params)
    print('test on Web Browser' + TTS_URL + '?' + data)

    req = Request(TTS_URL, data.encode('utf-8'))
    has_error = False
    try:
        f = urlopen(req)
        result_str = f.read()

        headers = dict((name.lower(), value) for name, value in f.headers.items())

        has_error = ('content-type' not in headers.keys() or headers['content-type'].find('audio/') < 0)
    except  URLError as err:
        print('asr http response http code : ' + str(err.code))
        result_str = err.read()
        has_error = True

    save_file = "error.txt" if has_error else 'result.' + FORMAT
    with open(save_file, 'wb') as of:
        of.write(result_str)

    if has_error:
        if (IS_PY3):
            result_str = str(result_str, 'utf-8')
        print("tts api  error:" + result_str)
    else:
        # 初始化pygame
        pygame.mixer.init()

        # 将音频数据加载到内存
        pygame.mixer.music.load(io.BytesIO(result_str))

        # 播放音频
        pygame.mixer.music.play()

        # # 防止程序立即退出
        # while pygame.mixer.music.get_busy():
        #     time.sleep(1)
    print("result saved as :" + save_file)