Python实现键盘按键触发多正弦波同时播放的问题求助

问题背景 #

编程基础薄弱，为大学研讨课开发Python按键触发发声程序，使用PyCharm实现，但当前按下"A"或"S"键只能依次播放对应音调，必须等待当前播放结束才能触发下一个，无法实现双音调同时叠加播放，需要修改代码实现多音同步输出。

原代码 #

import pyaudio
import numpy as np
import math
import struct
import keyboard

FS = 44100  #  frames per second, samples per second or sample rate

def play_sound(type, frequency, volume, duration):
   generate_sound(type, frequency, volume, duration)

def generate_sound(type, frequency, volume, duration):
    outbuf = np.random.normal(loc=0, scale=1, size=int(float(duration / 1000.0)*FS))

    if type == "sine":
        dur = int(FS * float(duration / 1000.0))
        theta = 0.0
        incr_theta = frequency * 2 * math.pi / FS # frequency increment normalized for sample rate
        for i in range(dur):
            outbuf[i] = volume * math.sin(theta)
            theta += incr_theta

    p = pyaudio.PyAudio()
    stream = p.open(format=pyaudio.paFloat32, channels=1, rate=FS, output=True)
    data = b''.join(struct.pack('f', samp) for samp in outbuf) # must pack the binary data
    stream.write(data)
    stream.stop_stream()
    stream.close()
    p.terminate()

#thread here maybe?

while True:
    if keyboard.is_pressed("a"):
        play_sound("sine",200, 0.5, 3000)
    if keyboard.is_pressed("s"):
        play_sound("sine",300,0.5,3000)
    elif keyboard.is_pressed("esc"):
        break

解决方案 #

原代码的核心问题是每次播放都创建独立的PyAudio实例和流，且stream.write()是阻塞操作，导致无法同时输出多音调。要实现多音叠加，需要改为单流持续输出+实时波形叠加的模式，具体修改如下：

修改后完整代码

import pyaudio
import math
import struct
import keyboard
import threading
import time

FS = 44100  # 采样率
VOLUME = 0.3  # 全局音量，避免叠加后爆音
CHUNK_SIZE = 1024  # 每次输出的音频块大小

# 保存当前激活的音符（频率）
active_notes = set()
# 每个音符的相位跟踪，避免切换时波形断层
note_phases = {}

def generate_sine_sample(frequency, phase):
    """生成单个采样点的正弦波值"""
    sample = VOLUME * math.sin(phase)
    # 更新相位
    new_phase = phase + (frequency * 2 * math.pi / FS)
    # 相位取模，防止数值过大
    return sample, new_phase % (2 * math.pi)

def audio_loop():
    """持续输出叠加后的音频流"""
    p = pyaudio.PyAudio()
    # 打开持续输出的流
    stream = p.open(
        format=pyaudio.paFloat32,
        channels=1,
        rate=FS,
        output=True,
        frames_per_buffer=CHUNK_SIZE
    )

    while True:
        chunk = []
        for _ in range(CHUNK_SIZE):
            sample = 0.0
            # 叠加所有激活音符的采样
            for freq in list(active_notes):
                # 初始化相位（如果是新音符）
                if freq not in note_phases:
                    note_phases[freq] = 0.0
                s, new_phase = generate_sine_sample(freq, note_phases[freq])
                sample += s
                note_phases[freq] = new_phase
            # 叠加后做限幅，防止音量过大失真
            sample = max(min(sample, 1.0), -1.0)
            chunk.append(sample)
        
        # 打包成二进制数据写入流
        data = b''.join(struct.pack('f', samp) for samp in chunk)
        stream.write(data)
        
        # 短暂休眠，降低CPU占用
        time.sleep(0.001)

def main():
    # 启动音频输出线程
    audio_thread = threading.Thread(target=audio_loop, daemon=True)
    audio_thread.start()

    # 按键状态跟踪，避免重复触发
    key_states = {"a": False, "s": False}

    print("按下A/S键播放音调，ESC键退出")
    while True:
        # 检测A键：200Hz
        if keyboard.is_pressed("a"):
            if not key_states["a"]:
                active_notes.add(200)
                key_states["a"] = True
        else:
            if key_states["a"]:
                active_notes.discard(200)
                note_phases.pop(200, None)
                key_states["a"] = False
        
        # 检测S键：300Hz
        if keyboard.is_pressed("s"):
            if not key_states["s"]:
                active_notes.add(300)
                key_states["s"] = True
        else:
            if key_states["s"]:
                active_notes.discard(300)
                note_phases.pop(300, None)
                key_states["s"] = False
        
        # 退出程序
        if keyboard.is_pressed("esc"):
            break

if __name__ == "__main__":
    main()

关键修改说明

• 线程分离：用threading把音频输出循环放到后台线程，主线程专注于按键检测，避免阻塞。
• 实时波形叠加：维护active_notes集合保存当前按下的音符，每次生成音频块时叠加所有激活音符的采样值。
• 相位跟踪：用note_phases字典记录每个音符的当前相位，避免按键松开再按下时波形出现断层。
• 限幅处理：叠加后的采样值做max(min(sample, 1.0), -1.0)处理，防止多个音调叠加导致音量过大失真。
• 按键状态跟踪：用key_states避免按键按住时重复添加音符到集合中。

内容的提问来源于stack exchange，提问作者BeaNatural