WAVE file read / write.

wav_io.py

# -*- coding: utf-8 -*-
"""
Created on 2020/02/13 

Description: WAVE file read / write.
@author: Jam Ytse
"""

import wave
from scipy import fromstring, int16
import numpy as np


def wave_read2ch(filename):
    """2ch WAVE file read.

    Extended description of function.

    Args:
        filename (str): The file name of wave file

    Returns:
        int: Frame rate of wav file.
        float: Normalized coef.
        ndarray: Left ch data read from wav file.
        ndarray: Right ch data read from wav file.

    """

    # Read sound data
    wav_obj = wave.open(filename, 'rb')

    # Get properties
    ch = wav_obj.getnchannels()
    width = wav_obj.getsampwidth()
    fr = wav_obj.getframerate()
    fn = wav_obj.getnframes()

    # Show properties
    print("Channel: ", ch)
    print("Sample width: ", width)
    print("Frame Rate: ", fr)
    print("Frame num: ", fn)
    print("Params: ", wav_obj.getparams())
    print("Total time: ", 1.0 * fn / fr)

    # Convert Wave_read object to bytes object
    data = wav_obj.readframes(wav_obj.getnframes())
    wav_data = fromstring(data, dtype=int16)
    # Seperate right and left ch
    if (wav_obj.getnchannels() == 2):
        # Left ch
        in_left = np.array(wav_data[::2])
        # Right ch
        in_right = np.array(wav_data[1::2])

    # Close wavw file
    wav_obj.close()

    # Convert int16 to float16
    _min = np.min(in_right)
    _max = np.max(in_right)
    max_right = max(abs(_min), abs(_max))
    _min = np.min(in_left)
    _max = np.max(in_left)
    max_left = max(abs(_min), abs(_max))
    # max_left = max(abs(in_left), abs(min(in_left)))
    max_total = max(max_right, max_left)

    # Normalize. Range -1.0 to +1.0
    in_left = in_left / max_total * 0.8
    in_right = in_right / max_total * 0.8

    return fr, max_total, in_left, in_right


def wave_write2ch(filename, fr, norm, left, right):
    """2ch WAVE File Write.

    Write a numpy array as a WAV file.

    Args:
        filename (str): The file name of output wave file.
        fr (int): The sample rate (in samples/sec).
        norm (float): Normalized coef.
        left(ndarray): The sound data of left channel.
        right(ndarray): The sound data of right channel.

    Returns:
        None.

    Note:
        Writes a simple uncompressed WAV file.
    """

    out_left = left * norm * 0.9
    out_right = right * norm * 0.9

    out = np.array([out_left, out_right]).transpose()
    out_data = out.astype(int16)

    wav_obj = wave.open(filename, 'wb')
    wav_obj.setnchannels(2)  # stereo
    wav_obj.setsampwidth(2)  # 16bits
    wav_obj.setframerate(fr)
    wav_obj.writeframes(out_data.tostring())
    wav_obj.close()