Skip to main content
 首页 » 编程设计

python之正确配置 Aubio/Alsaaudio

2025年01月19日7tuyile006

我正在尝试将 aubio 和 python 用于学校项目,目标是:检测何时有人发出两种声音,每种声音的长度为 2 秒,并且它们之间的间隔最大为 3 秒。第二个需要高于第一个。当满足这些条件时,程序会发送 Wake-On-Lan 包(当前代码中未实现)。

import alsaaudio 
import numpy as np 
import aubio 
import time 
import threading 
 
 
 
class Audio_watcher: 
    # constants 
    samplerate = 44100 
    win_s = 2048 
    hop_s = win_s // 2 
    framesize = hop_s 
    nb_samples = 20 
    tone_duration = 2.0 
    per_sampling = tone_duration / nb_samples 
    tone_max_interval = 3.0 
    tone_diff_ratio = 2 
 
 
    def __init__(self): 
        self.last_frequencies = np.zeros(Audio_watcher.nb_samples) 
        self.last_energies = np.zeros(Audio_watcher.nb_samples) 
        self.detected_tone = 0 
 
        # set up audio input 
        recorder = alsaaudio.PCM(type=alsaaudio.PCM_CAPTURE) 
        recorder.setperiodsize(Audio_watcher.framesize) 
        recorder.setrate(Audio_watcher.samplerate) 
        recorder.setformat(alsaaudio.PCM_FORMAT_FLOAT_LE) 
        recorder.setchannels(1) 
        self.recorder = recorder 
 
        pitcher = aubio.pitch("default", Audio_watcher.win_s, Audio_watcher.hop_s, Audio_watcher.samplerate) 
        pitcher.set_unit("Hz") 
        pitcher.set_silence(-40) 
        self.pitcher = pitcher 
        # A filter 
        f = aubio.digital_filter(7) 
        f.set_a_weighting(Audio_watcher.samplerate) 
        self.f = f 
 
 
    def get_audio(self): 
        # read and convert data from audio input 
        _, data = self.recorder.read() 
        samples = np.fromstring(data, dtype=aubio.float_type) 
        filtered_samples = self.f(samples) 
        print(filtered_samples) 
 
        # pitch and energy of current frame 
        freq = self.pitcher(filtered_samples)[0] 
        print(freq) 
        self.last_frequencies = np.roll(self.last_frequencies, 1) 
        self.last_frequencies[0] = freq 
        self.last_energies = np.roll(self.last_energies, 1) 
        self.last_energies[0] = np.sum(filtered_samples**2)/len(filtered_samples) 
 
        threading.Timer(Audio_watcher.per_sampling, self.get_audio).start() 
 
 
    def reset_detected_tone(): 
        self.detected_tone = 0 
 
 
    def detect_tone(self): 
        std_last = np.std(self.last_frequencies) 
        if std_last <= 200 and std_last > 0: 
            mean_freq = np.mean(self.last_frequencies) 
            if self.detected_tone == 0: 
                self.detected_tone = mean_freq 
                threading.Timer(Audio_watcher.tone_max_interval, self.reset_detected_tone).start() 
            elif mean_freq > Audio_watcher.tone_diff_ratio * self.detected_tone: 
                print('wol') 
 
        threading.Timer(Audio_watcher.tone_duration, self.detect_tone).start() 
 
 
 
aw = Audio_watcher() 
aw.get_audio() 
aw.detect_tone() 

然而,使用这段代码,我在声音和它们的检测之间有很大的延迟,我认为这与每 0.1 秒只调用一次录音机有关,但我找不到如何为 aubio 提供正确的参数.
有谁知道如何配置常量以使其工作?
非常感谢!

请您参考如下方法:

找出导致此错误的原因,我需要将设置音频输入的代码放在 get_audio 函数中,以便它每次都更新