webassembly003 whisper.cpp的python绑定实现+Cython+Setuptools的GUI程序

CODE

• python端的绑定和本文一样，还需要将cdef char* LANGUAGE = b’en’改为中文zh（也可以在函数中配置一个参数修改这个值）。
• ps:本来想尝试cdef whisper_context* whisper_init_from_file_with_params_no_state(char*, whisper_full_params)然后进行调用，但是发现最新版的whisper.h没有这个API了，所以先不加了。

import pyaudio
import wave
import struct
import sys
import numpy as np

import pyqtgraph as pg
from PyQt5 import QtWidgets
from PyQt5.QtCore import Qt

from whispercpp import Whisper

# Audio Format (check Audio MIDI Setup if on Mac)
FORMAT = pyaudio.paInt16
RATE = 16000
CHANNELS = 2

# Set Plot Range [-RANGE,RANGE], default is nyquist/2
RANGE = None
if not RANGE:
	RANGE = RATE/2

# Set these parameters (How much data to plot per FFT)
INPUT_BLOCK_TIME = 0.05
INPUT_FRAMES_PER_BLOCK = int(RATE*INPUT_BLOCK_TIME)

# Which Channel? (L or R)
LR = "l"

class SpectrumAnalyzer():
	def __init__(self):
		self.pa = pyaudio.PyAudio()
		self.initMicrophone()
		self.initUI()

	def find_input_device(self):
		device_index = None            
		for i in range(self.pa.get_device_count()):     
			devinfo = self.pa.get_device_info_by_index(i)
			if devinfo["name"].lower() in ["mic","input"]:
				device_index = i
		return device_index

	def initMicrophone(self):
		device_index = self.find_input_device()

		self.stream = self.pa.open(	format = FORMAT,
									channels = CHANNELS,
									rate = RATE,
									input = True,
									input_device_index = device_index,
									frames_per_buffer = INPUT_FRAMES_PER_BLOCK)

	def readData(self):
		block = self.stream.read(INPUT_FRAMES_PER_BLOCK)
		count = len(block)/2
		format = "%dh"%(count)
		shorts = struct.unpack( format, block )
		if CHANNELS == 1:
			return np.array(shorts)
		else:
			l = shorts[::2]
			r = shorts[1::2]
			if LR == 'l':
				return np.array(l)
			else:
				return np.array(r)

	def initUI(self):
		self.app = QtWidgets.QApplication([]) # self.app = QtGui.QApplication([])
		self.app.quitOnLastWindowClosed()

		self.mainWindow = QtWidgets.QMainWindow()
		self.mainWindow.setWindowFlags(Qt.FramelessWindowHint | Qt.WindowStaysOnTopHint)
		self.mainWindow.setWindowTitle("Spectrum Analyzer")
		self.mainWindow.setGeometry(100, 100, 300, 200)#self.mainWindow.resize(800,300)
		self.centralWid = QtWidgets.QWidget()
		self.mainWindow.setCentralWidget(self.centralWid)
		self.lay = QtWidgets.QVBoxLayout()
		self.centralWid.setLayout(self.lay)
  
		# Add a button
		self.button_start = QtWidgets.QPushButton("Start Record Audio")
		self.button_start.clicked.connect(self.Button_Start)
		self.lay.addWidget(self.button_start)
		self.button_end = QtWidgets.QPushButton("whisper Init")
		self.whisper = None
		self.is_whisper_inited = False
		self.button_end.clicked.connect(self.Button_Whisper)
		self.lay.addWidget(self.button_end)
		self.button = QtWidgets.QPushButton("TRANS AUDIO")
		self.button.clicked.connect(self.Button_TransAudio)
		self.lay.addWidget(self.button)
		# Add a text label
		self.label = QtWidgets.QLabel("Text will appear here:")
		self.lay.addWidget(self.label)
        # Add a QLineEdit
		self.text_field = QtWidgets.QLineEdit()
		self.text_field.setFixedSize(280, 200)
		self.lay.addWidget(self.text_field)
  
		self.specWid = pg.PlotWidget(name="spectrum")
		self.specItem = self.specWid.getPlotItem()
		self.specItem.setMouseEnabled(y=False)
		self.specItem.setYRange(0,1000)
		self.specItem.setXRange(-RANGE,RANGE, padding=0)

		self.specAxis = self.specItem.getAxis("bottom")
		self.specAxis.setLabel("Frequency [Hz]")
		self.lay.addWidget(self.specWid)

		self.mainWindow.show()
		self.app.aboutToQuit.connect(self.close)
  
	def onButtonClick(self):
		self.label.setText("Whisper res is:")
		self.text_field.setText("Hello")
  
	def Button_Whisper(self):
		self.whisper = Whisper('large',model_path= "/home/pdd/myassets/ggml-medium.bin")
		self.is_whisper_inited = True
		self.text_field.setText("Whisper INITED")
  
	def Button_TransAudio(self):
		result = self.whisper.transcribe("/home/pdd/le/pywhisper/output.wav") # result = w.transcribe("myfile.mp3")
		print(123)
		text = self.whisper.extract_text(result)
		self.text_field.setText(str(text))
  
	def Button_Start(self):
		self.label.setText("Whisper res is:")
		self.text_field.setText("Start ---")
		# 录制音频
		frames = []
		sample_rate = 16000
		duration = 5
		for i in range(0, int(sample_rate / 1024 * duration)):
			data = self.stream.read(1024)
			frames.append(data)
   
		# 将录制的音频保存为wav文件
		with wave.open("output.wav", 'wb') as wf:
			wf.setnchannels(CHANNELS) # 2
			wf.setsampwidth(self.pa.get_sample_size(FORMAT)) # 2
			wf.setframerate(sample_rate)
			wf.writeframes(b''.join(frames))
		self.text_field.setText("保存为wav文件")
		

	def close(self):
		self.stream.close()
		sys.exit()

	def get_spectrum(self, data):
		T = 1.0/RATE
		N = data.shape[0]
		Pxx = (1./N)*np.fft.fft(data)
		f = np.fft.fftfreq(N,T)
		Pxx = np.fft.fftshift(Pxx)
		f = np.fft.fftshift(f)

		return f.tolist(), (np.absolute(Pxx)).tolist()

	def mainLoop(self):
		while 1:
			# Sometimes Input overflowed because of mouse events, ignore this
			try:
				data = self.readData()
			except IOError:
				continue
			f, Pxx = self.get_spectrum(data)
			self.specItem.plot(x=f,y=Pxx, clear=True)
			QtWidgets.QApplication.processEvents()

if __name__ == '__main__':
	sa = SpectrumAnalyzer()
	sa.mainLoop()