使用speech_recognition库报错：audio_data需为合法音频数据的问题

问题根源 #

soundcard录制返回的data是numpy浮点数组，而speech_recognition的recognize_google方法仅接受该库封装的AudioData对象，或通过AudioFile/Microphone读取的音频源，直接传numpy数组会触发类型不匹配报错。

解决方案1：将numpy数组转换为AudioData对象 #

直接处理录制的数组，转换成speech_recognition要求的格式：

import numpy as np
import soundcard as sc
import speech_recognition as sr

OUTPUT_FILE_NAME = "out.wav"    # 文件名
SAMPLE_RATE = 48000              # 采样率[Hz]
RECORD_SEC = 5                  # 录制时长[秒]

# 录制扬声器回环音频
with sc.get_microphone(id=str(sc.default_speaker().name), include_loopback=True).recorder(samplerate=SAMPLE_RATE) as mic:
    data = mic.record(numframes=SAMPLE_RATE * RECORD_SEC)

# 适配speech_recognition的音频格式要求
# 1. 立体声转单声道（识别无需立体声，取声道平均值）
if data.ndim > 1:
    data = np.mean(data, axis=1)
# 2. 将float32格式（范围[-1,1]）转为int16格式（PCM标准，范围[-32768,32767]）
data_int16 = (data * 32767).astype(np.int16)
# 3. 创建符合要求的AudioData对象
audio_data = sr.AudioData(data_int16.tobytes(), sample_rate=SAMPLE_RATE, sample_width=2)

# 调用Google语音识别
r = sr.Recognizer()
try:
    text = r.recognize_google(audio_data, language='en-US', show_all=False, key=None)
    print(text)
except sr.UnknownValueError:
    print("Google语音识别无法理解音频内容")
except sr.RequestError as e:
    print("无法从Google语音识别服务获取结果；{0}".format(e))

解决方案2：保存为WAV文件后读取识别 #

先把录制的音频存为标准WAV文件，再用speech_recognition加载识别：

import soundcard as sc
import soundfile as sf
import speech_recognition as sr

OUTPUT_FILE_NAME = "out.wav"    # 文件名
SAMPLE_RATE = 48000              # 采样率[Hz]
RECORD_SEC = 5                  # 录制时长[秒]

# 录制并保存为WAV文件
with sc.get_microphone(id=str(sc.default_speaker().name), include_loopback=True).recorder(samplerate=SAMPLE_RATE) as mic:
    data = mic.record(numframes=SAMPLE_RATE * RECORD_SEC)
sf.write(OUTPUT_FILE_NAME, data, SAMPLE_RATE)

# 读取WAV文件并执行识别
r = sr.Recognizer()
with sr.AudioFile(OUTPUT_FILE_NAME) as source:
    audio_data = r.record(source)
try:
    text = r.recognize_google(audio_data, language='en-US', show_all=False, key=None)
    print(text)
except sr.UnknownValueError:
    print("Google语音识别无法理解音频内容")
except sr.RequestError as e:
    print("无法从Google语音识别服务获取结果；{0}".format(e))

内容的提问来源于stack exchange，提问作者ProfessionalAli