树莓派3项目：获取WAV文件FFT遇阻，寻求技术解决方案

Hey there, let's work through the FFT problem you're hitting with your Raspberry Pi 3 audio project. Looking at your code, there are a few clear issues that are preventing you from getting a valid FFT of your doorbell signal—let's break them down and fix it step by step.

First, Let's Spot the Code Problems #

Your current code has two critical mistakes in how you're handling the WAV file data:

• You convert input_data[1] (the audio samples) to a float array but don't assign it to any variable—so that line does nothing.
• The line audio = input_data[...] is invalid syntax and doesn't correctly capture the audio samples from the WAV file's output.

Corrected Code with Explanations #

Here's a revised version of your code that properly reads the audio, computes the FFT, and plots the result:

import scipy
from scipy.io.wavfile import read
from scipy.signal import hann
from scipy.fftpack import rfft, rfftfreq
import matplotlib.pyplot as plt
import numpy as np

# Read the WAV file: returns (sample_rate, audio_samples)
sample_rate, raw_audio = read("/home/pi/Music/doorbell-1.wav")

# Convert raw integer audio samples to normalized floats (range -1.0 to 1.0)
# Most WAV files use 16-bit integers, so divide by 32768.0
audio = np.array(raw_audio, dtype=np.float32) / 32768.0

# Handle stereo audio: if the audio has 2 channels, convert to mono by averaging
if len(audio.shape) > 1:
    audio = np.mean(audio, axis=1)

# Apply Hann window to reduce spectral leakage
window = hann(len(audio))
windowed_audio = audio * window

# Compute real FFT (since audio is real-valued, rfft is more efficient)
fft_result = rfft(windowed_audio)

# Calculate corresponding frequency values for the FFT bins
freqs = rfftfreq(len(windowed_audio), 1/sample_rate)

# Compute magnitude spectrum (convert complex values to absolute magnitude)
magnitude = np.abs(fft_result)

# Plot the FFT result
plt.figure(figsize=(10, 6))
plt.plot(freqs, magnitude)
plt.xlabel("Frequency (Hz)")
plt.ylabel("Magnitude")
plt.title("FFT of Doorbell Audio Signal")
plt.grid(True)
plt.show()

Key Fixes & Notes #

• Proper WAV Data Handling: We explicitly capture the sample_rate and raw_audio from the read() function, then normalize the integer samples to floats—this is crucial because FFT works best with floating-point values.
• Stereo to Mono Conversion: Many WAV files are stereo; averaging the channels ensures you're working with a single signal for your FFT.
• Windowing: Applying the Hann window reduces spectral leakage (artificial spread of energy across frequencies), which makes your FFT results more accurate.
• Frequency Axis Calculation: Using rfftfreq automatically computes the correct frequency bins for the real FFT, so you don't have to manually calculate them.
• Magnitude Spectrum: The FFT returns complex numbers, so taking the absolute value gives you the magnitude of each frequency component.

Next Steps for Your Bandpass Filter #

Once you have a valid FFT to verify your target signal's frequency range, you can use scipy.signal.butter or scipy.signal.firwin to design your bandpass filter, then apply it with scipy.signal.lfilter.

内容的提问来源于stack exchange，提问作者nanriat