Splitting PDF Files and Recognizing MP3 Audio with Python

In this tutorial we demonstrate how to programmatically split a PDF document into separate files—extracting the first page as a cover and grouping subsequent pages in pairs—using the PyPDF2 library.

We then show two approaches for processing an accompanying MP3 audio file: first converting the MP3 to WAV and using the Google Speech‑Recognition API for accurate but potentially incomplete transcription, and second using PocketSphinx for a full transcription that may be less accurate.

Both methods are implemented in Python, with complete code examples provided below.

from PyPDF2 import PdfReader, PdfWriter

def split_pdf(input_path, output_folder):
    with open(input_path, 'rb') as file:
        reader = PdfReader(file)
        total_pages = len(reader.pages)

        # Split the first page as a separate PDF
        first_page_output_path = f"{output_folder}/1.pdf"
        first_page_writer = PdfWriter()
        first_page = reader.pages[0]
        first_page_writer.add_page(first_page)
        with open(first_page_output_path, 'wb') as first_page_output_file:
            first_page_writer.write(first_page_output_file)

        # Split the remaining pages
        for start_page in range(1, total_pages, 2):
            end_page = min(start_page + 1, total_pages - 1)
            output_path = f"{output_folder}/{start_page + 1}-{end_page + 1}.pdf"
            writer = PdfWriter()
            for page_number in range(start_page, end_page + 1):
                page = reader.pages[page_number]
                writer.add_page(page)
            with open(output_path, 'wb') as output_file:
                writer.write(output_file)

input_pdf_path = 'XXX.pdf'
output_folder = '/output_folder/pdf'

if __name__ == '__main__':
    split_pdf(input_pdf_path, output_folder)

import os
import speech_recognition as sr
from pydub import AudioSegment

def mp3_to_wav(input_path, output_path):
    # 读取MP3文件
    audio = AudioSegment.from_mp3(input_path)
    # 将MP3转换为WAV格式
    audio.export(output_path, format="wav")

def recognize_speech(input_wav_path, output_txt_path):
    # 初始化语音识别器
    recognizer = sr.Recognizer()
    # 读取WAV文件并进行语音识别
    with sr.AudioFile(input_wav_path) as source:
        audio = recognizer.record(source)
    # 进行语音识别
    try:
        text = recognizer.recognize_google(audio, language="en-US")
    except sr.UnknownValueError:
        text = "Speech Recognition could not understand the audio."
    except sr.RequestError as e:
        text = f"Could not request results from Google Speech Recognition service; {e}"
    # 将识别结果保存到文本文件中
    with open(output_txt_path, "a", encoding="utf-8") as output_file:
        output_file.write(text)

if __name__ == "__main__":
    input_mp3 = "XXX.mp3"
    output_wav = "/output_folder/output1.wav"
    mp3_to_wav(input_mp3, output_wav)
    output_txt = "/output_folder/output1.txt"
    recognize_speech(output_wav, output_txt)

import speech_recognition as sr
from pydub import AudioSegment

# 定义输入和输出文件路径
input_file = "XXX.mp3"
output_file = "/output_folder/output.txt"

# 将 MP3 文件转换为 WAV 格式
audio = AudioSegment.from_mp3(input_file)
audio.export("/output_folder/temp.wav", format="wav")

# 创建语音识别器对象
recognizer = sr.Recognizer()

# 使用 PocketSphinx 进行识别
with sr.AudioFile("/output_folder/temp.wav") as source:
    audio_data = recognizer.record(source)
    text = recognizer.recognize_sphinx(audio_data)

# 将识别的文本按行写入文件
lines = text.split("\n")
with open(output_file, "w") as file:
    for line in lines:
        file.write(line + "\n")

print("转换完成！")