cros/audio/audio_helper.py

#!/usr/bin/python
# Copyright (c) 2012 The Chromium OS Authors. All rights reserved.
# Use of this source code is governed by a BSD-style license that can be
# found in the LICENSE file.


import logging
import numpy
import os
import re
import subprocess
import tempfile
import threading
import time

from glob import glob
from autotest_lib.client.bin import test, utils
from autotest_lib.client.bin.input.input_device import *
from autotest_lib.client.common_lib import error
from autotest_lib.client.cros.audio import audio_data
from autotest_lib.client.cros.audio import cmd_utils
from autotest_lib.client.cros.audio import cras_utils
from autotest_lib.client.cros.audio import sox_utils

LD_LIBRARY_PATH = 'LD_LIBRARY_PATH'

_AUDIO_DIAGNOSTICS_PATH = '/usr/bin/audio_diagnostics'

_DEFAULT_NUM_CHANNELS = 2
_DEFAULT_REC_COMMAND = 'arecord -D hw:0,0 -d 10 -f dat'
_DEFAULT_SOX_FORMAT = '-t raw -b 16 -e signed -r 48000 -L'
_DEFAULT_PLAYBACK_VOLUME = 100
_DEFAULT_CAPTURE_GAIN = 2500
_DEFAULT_ALSA_MAX_VOLUME = '100%'
_DEFAULT_ALSA_CAPTURE_GAIN = '25dB'

# Minimum RMS value to pass when checking recorded file.
_DEFAULT_SOX_RMS_THRESHOLD = 0.08

_JACK_VALUE_ON_RE = re.compile(r'.*values=on')
_HP_JACK_CONTROL_RE = re.compile(r'numid=(\d+).*Headphone\sJack')
_MIC_JACK_CONTROL_RE = re.compile(r'numid=(\d+).*Mic\sJack')

_SOX_RMS_AMPLITUDE_RE = re.compile(r'RMS\s+amplitude:\s+(.+)')
_SOX_ROUGH_FREQ_RE = re.compile(r'Rough\s+frequency:\s+(.+)')

_AUDIO_NOT_FOUND_RE = r'Audio\snot\sdetected'
_MEASURED_LATENCY_RE = r'Measured\sLatency:\s(\d+)\suS'
_REPORTED_LATENCY_RE = r'Reported\sLatency:\s(\d+)\suS'

# Tools from platform/audiotest
AUDIOFUNTEST_PATH = 'audiofuntest'
AUDIOLOOP_PATH = 'looptest'
LOOPBACK_LATENCY_PATH = 'loopback_latency'
SOX_PATH = 'sox'
TEST_TONES_PATH = 'test_tones'

_MINIMUM_NORM = 0.001
_CORRELATION_INDEX_THRESHOLD = 0.999
# The minimum difference of estimated frequencies between two sine waves.
_FREQUENCY_DIFF_THRESHOLD = 20
# The minimum RMS value of meaningful audio data.
_MEANINGFUL_RMS_THRESHOLD = 0.001

def set_mixer_controls(mixer_settings={}, card='0'):
    """Sets all mixer controls listed in the mixer settings on card.

    @param mixer_settings: Mixer settings to set.
    @param card: Index of audio card to set mixer settings for.
    """
    logging.info('Setting mixer control values on %s', card)
    for item in mixer_settings:
        logging.info('Setting %s to %s on card %s',
                     item['name'], item['value'], card)
        cmd = 'amixer -c %s cset name=%s %s'
        cmd = cmd % (card, item['name'], item['value'])
        try:
            utils.system(cmd)
        except error.CmdError:
            # A card is allowed not to support all the controls, so don't
            # fail the test here if we get an error.
            logging.info('amixer command failed: %s', cmd)

def set_volume_levels(volume, capture):
    """Sets the volume and capture gain through cras_test_client.

    @param volume: The playback volume to set.
    @param capture: The capture gain to set.
    """
    logging.info('Setting volume level to %d', volume)
    try:
        utils.system('/usr/bin/cras_test_client --volume %d' % volume)
        logging.info('Setting capture gain to %d', capture)
        utils.system('/usr/bin/cras_test_client --capture_gain %d' % capture)
        utils.system('/usr/bin/cras_test_client --dump_server_info')
        utils.system('/usr/bin/cras_test_client --mute 0')
    except error.CmdError as e:
        raise error.TestError(
                '*** Can not tune volume through CRAS. *** (' + str(e) + ')')

    try:
        utils.system('amixer -c 0 contents')
    except error.CmdError as e:
        logging.info('amixer command failed: %s', str(e))

def loopback_latency_check(**args):
    """Checks loopback latency.

    @param args: additional arguments for loopback_latency.

    @return A tuple containing measured and reported latency in uS.
        Return None if no audio detected.
    """
    noise_threshold = str(args['n']) if args.has_key('n') else '400'

    cmd = '%s -n %s -c' % (LOOPBACK_LATENCY_PATH, noise_threshold)

    output = utils.system_output(cmd, retain_output=True)

    # Sleep for a short while to make sure device is not busy anymore
    # after called loopback_latency.
    time.sleep(.1)

    measured_latency = None
    reported_latency = None
    for line in output.split('\n'):
        match = re.search(_MEASURED_LATENCY_RE, line, re.I)
        if match:
            measured_latency = int(match.group(1))
            continue
        match = re.search(_REPORTED_LATENCY_RE, line, re.I)
        if match:
            reported_latency = int(match.group(1))
            continue
        if re.search(_AUDIO_NOT_FOUND_RE, line, re.I):
            return None
    if measured_latency and reported_latency:
        return (measured_latency, reported_latency)
    else:
        # Should not reach here, just in case.
        return None

def get_mixer_jack_status(jack_reg_exp):
    """Gets the mixer jack status.

    @param jack_reg_exp: The regular expression to match jack control name.

    @return None if the control does not exist, return True if jack control
        is detected plugged, return False otherwise.
    """
    output = utils.system_output('amixer -c0 controls', retain_output=True)
    numid = None
    for line in output.split('\n'):
        m = jack_reg_exp.match(line)
        if m:
            numid = m.group(1)
            break

    # Proceed only when matched numid is not empty.
    if numid:
        output = utils.system_output('amixer -c0 cget numid=%s' % numid)
        for line in output.split('\n'):
            if _JACK_VALUE_ON_RE.match(line):
                return True
        return False
    else:
        return None

def get_hp_jack_status():
    """Gets the status of headphone jack."""
    status = get_mixer_jack_status(_HP_JACK_CONTROL_RE)
    if status is not None:
        return status

    # When headphone jack is not found in amixer, lookup input devices
    # instead.
    #
    # TODO(hychao): Check hp/mic jack status dynamically from evdev. And
    # possibly replace the existing check using amixer.
    for evdev in glob('/dev/input/event*'):
        device = InputDevice(evdev)
        if device.is_hp_jack():
            return device.get_headphone_insert()
    else:
        return None

def get_mic_jack_status():
    """Gets the status of mic jack."""
    status = get_mixer_jack_status(_MIC_JACK_CONTROL_RE)
    if status is not None:
        return status

    # When mic jack is not found in amixer, lookup input devices instead.
    for evdev in glob('/dev/input/event*'):
        device = InputDevice(evdev)
        if device.is_mic_jack():
            return device.get_microphone_insert()
    else:
        return None

def log_loopback_dongle_status():
    """Log the status of the loopback dongle to make sure it is equipped."""
    dongle_status_ok = True

    # Check Mic Jack
    mic_jack_status = get_mic_jack_status()
    logging.info('Mic jack status: %s', mic_jack_status)
    dongle_status_ok &= bool(mic_jack_status)

    # Check Headphone Jack
    hp_jack_status = get_hp_jack_status()
    logging.info('Headphone jack status: %s', hp_jack_status)
    dongle_status_ok &= bool(hp_jack_status)

    # Use latency check to test if audio can be captured through dongle.
    # We only want to know the basic function of dongle, so no need to
    # assert the latency accuracy here.
    latency = loopback_latency_check(n=4000)
    if latency:
        logging.info('Got latency measured %d, reported %d',
                latency[0], latency[1])
    else:
        logging.info('Latency check fail.')
        dongle_status_ok = False

    logging.info('audio loopback dongle test: %s',
            'PASS' if dongle_status_ok else 'FAIL')

# Functions to test audio palyback.
def play_sound(duration_seconds=None, audio_file_path=None):
    """Plays a sound file found at |audio_file_path| for |duration_seconds|.

    If |audio_file_path|=None, plays a default audio file.
    If |duration_seconds|=None, plays audio file in its entirety.

    @param duration_seconds: Duration to play sound.
    @param audio_file_path: Path to the audio file.
    """
    if not audio_file_path:
        audio_file_path = '/usr/local/autotest/cros/audio/sine440.wav'
    duration_arg = ('-d %d' % duration_seconds) if duration_seconds else ''
    utils.system('aplay %s %s' % (duration_arg, audio_file_path))

def get_play_sine_args(channel, odev='default', freq=1000, duration=10,
                       sample_size=16):
    """Gets the command args to generate a sine wav to play to odev.

    @param channel: 0 for left, 1 for right; otherwize, mono.
    @param odev: alsa output device.
    @param freq: frequency of the generated sine tone.
    @param duration: duration of the generated sine tone.
    @param sample_size: output audio sample size. Default to 16.
    """
    cmdargs = [SOX_PATH, '-b', str(sample_size), '-n', '-t', 'alsa',
               odev, 'synth', str(duration)]
    if channel == 0:
        cmdargs += ['sine', str(freq), 'sine', '0']
    elif channel == 1:
        cmdargs += ['sine', '0', 'sine', str(freq)]
    else:
        cmdargs += ['sine', str(freq)]

    return cmdargs

def play_sine(channel, odev='default', freq=1000, duration=10,
              sample_size=16):
    """Generates a sine wave and plays to odev.

    @param channel: 0 for left, 1 for right; otherwize, mono.
    @param odev: alsa output device.
    @param freq: frequency of the generated sine tone.
    @param duration: duration of the generated sine tone.
    @param sample_size: output audio sample size. Default to 16.
    """
    cmdargs = get_play_sine_args(channel, odev, freq, duration, sample_size)
    utils.system(' '.join(cmdargs))

# Functions to compose customized sox command, execute it and process the
# output of sox command.
def get_sox_mixer_cmd(infile, channel,
                      num_channels=_DEFAULT_NUM_CHANNELS,
                      sox_format=_DEFAULT_SOX_FORMAT):
    """Gets sox mixer command to reduce channel.

    @param infile: Input file name.
    @param channel: The selected channel to take effect.
    @param num_channels: The number of total channels to test.
    @param sox_format: Format to generate sox command.
    """
    # Build up a pan value string for the sox command.
    if channel == 0:
        pan_values = '1'
    else:
        pan_values = '0'
    for pan_index in range(1, num_channels):
        if channel == pan_index:
            pan_values = '%s%s' % (pan_values, ',1')
        else:
            pan_values = '%s%s' % (pan_values, ',0')

    return '%s -c 2 %s %s -c 1 %s - mixer %s' % (SOX_PATH,
            sox_format, infile, sox_format, pan_values)

def sox_stat_output(infile, channel,
                    num_channels=_DEFAULT_NUM_CHANNELS,
                    sox_format=_DEFAULT_SOX_FORMAT):
    """Executes sox stat command.

    @param infile: Input file name.
    @param channel: The selected channel.
    @param num_channels: The number of total channels to test.
    @param sox_format: Format to generate sox command.

    @return The output of sox stat command
    """
    sox_mixer_cmd = get_sox_mixer_cmd(infile, channel,
                                      num_channels, sox_format)
    stat_cmd = '%s -c 1 %s - -n stat 2>&1' % (SOX_PATH, sox_format)
    sox_cmd = '%s | %s' % (sox_mixer_cmd, stat_cmd)
    return utils.system_output(sox_cmd, retain_output=True)

def get_audio_rms(sox_output):
    """Gets the audio RMS value from sox stat output

    @param sox_output: Output of sox stat command.

    @return The RMS value parsed from sox stat output.
    """
    for rms_line in sox_output.split('\n'):
        m = _SOX_RMS_AMPLITUDE_RE.match(rms_line)
        if m is not None:
            return float(m.group(1))

def get_rough_freq(sox_output):
    """Gets the rough audio frequency from sox stat output

    @param sox_output: Output of sox stat command.

    @return The rough frequency value parsed from sox stat output.
    """
    for rms_line in sox_output.split('\n'):
        m = _SOX_ROUGH_FREQ_RE.match(rms_line)
        if m is not None:
            return int(m.group(1))

def check_audio_rms(sox_output, sox_threshold=_DEFAULT_SOX_RMS_THRESHOLD):
    """Checks if the calculated RMS value is expected.

    @param sox_output: The output from sox stat command.
    @param sox_threshold: The threshold to test RMS value against.

    @raises error.TestError if RMS amplitude can't be parsed.
    @raises error.TestFail if the RMS amplitude of the recording isn't above
            the threshold.
    """
    rms_val = get_audio_rms(sox_output)

    # In case we don't get a valid RMS value.
    if rms_val is None:
        raise error.TestError(
            'Failed to generate an audio RMS value from playback.')

    logging.info('Got audio RMS value of %f. Minimum pass is %f.',
                 rms_val, sox_threshold)
    if rms_val < sox_threshold:
        raise error.TestFail(
            'Audio RMS value %f too low. Minimum pass is %f.' %
            (rms_val, sox_threshold))

def noise_reduce_file(in_file, noise_file, out_file,
                      sox_format=_DEFAULT_SOX_FORMAT):
    """Runs the sox command to reduce noise.

    Runs the sox command to noise-reduce in_file using the noise
    profile from noise_file.

    @param in_file: The file to noise reduce.
    @param noise_file: The file containing the noise profile.
        This can be created by recording silence.
    @param out_file: The file contains the noise reduced sound.
    @param sox_format: The  sox format to generate sox command.
    """
    prof_cmd = '%s -c 2 %s %s -n noiseprof' % (SOX_PATH,
               sox_format, noise_file)
    reduce_cmd = ('%s -c 2 %s %s -c 2 %s %s noisered' %
            (SOX_PATH, sox_format, in_file, sox_format, out_file))
    utils.system('%s | %s' % (prof_cmd, reduce_cmd))

def record_sample(tmpfile, record_command=_DEFAULT_REC_COMMAND):
    """Records a sample from the default input device.

    @param tmpfile: The file to record to.
    @param record_command: The command to record audio.
    """
    utils.system('%s %s' % (record_command, tmpfile))

def create_wav_file(wav_dir, prefix=""):
    """Creates a unique name for wav file.

    The created file name will be preserved in autotest result directory
    for future analysis.

    @param wav_dir: The directory of created wav file.
    @param prefix: specified file name prefix.
    """
    filename = "%s-%s.wav" % (prefix, time.time())
    return os.path.join(wav_dir, filename)

def run_in_parallel(*funs):
    """Runs methods in parallel.

    @param funs: methods to run.
    """
    threads = []
    for f in funs:
        t = threading.Thread(target=f)
        t.start()
        threads.append(t)

    for t in threads:
        t.join()

def loopback_test_channels(noise_file_name, wav_dir,
                           playback_callback=None,
                           check_recorded_callback=check_audio_rms,
                           preserve_test_file=True,
                           num_channels = _DEFAULT_NUM_CHANNELS,
                           record_callback=record_sample,
                           mix_callback=None):
    """Tests loopback on all channels.

    @param noise_file_name: Name of the file contains pre-recorded noise.
    @param wav_dir: The directory of created wav file.
    @param playback_callback: The callback to do the playback for
        one channel.
    @param record_callback: The callback to do the recording.
    @param check_recorded_callback: The callback to check recorded file.
    @param preserve_test_file: Retain the recorded files for future debugging.
    @param num_channels: The number of total channels to test.
    @param mix_callback: The callback to do on the one-channel file.
    """
    for channel in xrange(num_channels):
        record_file_name = create_wav_file(wav_dir,
                                           "record-%d" % channel)
        functions = [lambda: record_callback(record_file_name)]

        if playback_callback:
            functions.append(lambda: playback_callback(channel))

        if mix_callback:
            mix_file_name = create_wav_file(wav_dir, "mix-%d" % channel)
            functions.append(lambda: mix_callback(mix_file_name))

        run_in_parallel(*functions)

        if mix_callback:
            sox_output_mix = sox_stat_output(mix_file_name, channel)
            rms_val_mix = get_audio_rms(sox_output_mix)
            logging.info('Got mixed audio RMS value of %f.', rms_val_mix)

        sox_output_record = sox_stat_output(record_file_name, channel)
        rms_val_record = get_audio_rms(sox_output_record)
        logging.info('Got recorded audio RMS value of %f.', rms_val_record)

        reduced_file_name = create_wav_file(wav_dir,
                                            "reduced-%d" % channel)
        noise_reduce_file(record_file_name, noise_file_name,
                          reduced_file_name)

        sox_output_reduced = sox_stat_output(reduced_file_name, channel)

        if not preserve_test_file:
            os.unlink(reduced_file_name)
            os.unlink(record_file_name)
            if mix_callback:
                os.unlink(mix_file_name)

        check_recorded_callback(sox_output_reduced)


def get_channel_sox_stat(
        input_audio, channel_index, channels=2, bits=16, rate=48000):
    """Gets the sox stat info of the selected channel in the input audio file.

    @param input_audio: The input audio file to be analyzed.
    @param channel_index: The index of the channel to be analyzed.
                          (1 for the first channel).
    @param channels: The number of channels in the input audio.
    @param bits: The number of bits of each audio sample.
    @param rate: The sampling rate.
    """
    if channel_index <= 0 or channel_index > channels:
        raise ValueError('incorrect channel_indexi: %d' % channel_index)

    if channels == 1:
        return sox_utils.get_stat(
                input_audio, channels=channels, bits=bits, rate=rate)

    p1 = cmd_utils.popen(
            sox_utils.extract_channel_cmd(
                    input_audio, '-', channel_index,
                    channels=channels, bits=bits, rate=rate),
            stdout=subprocess.PIPE)
    p2 = cmd_utils.popen(
            sox_utils.stat_cmd('-', channels=1, bits=bits, rate=rate),
            stdin=p1.stdout, stderr=subprocess.PIPE)
    stat_output = p2.stderr.read()
    cmd_utils.wait_and_check_returncode(p1, p2)
    return sox_utils.parse_stat_output(stat_output)


def get_rms(input_audio, channels=1, bits=16, rate=48000):
    """Gets the RMS values of all channels of the input audio.

    @param input_audio: The input audio file to be checked.
    @param channels: The number of channels in the input audio.
    @param bits: The number of bits of each audio sample.
    @param rate: The sampling rate.
    """
    stats = [get_channel_sox_stat(
            input_audio, i + 1, channels=channels, bits=bits,
            rate=rate) for i in xrange(channels)]

    logging.info('sox stat: %s', [str(s) for s in stats])
    return [s.rms for s in stats]


def reduce_noise_and_get_rms(
        input_audio, noise_file, channels=1, bits=16, rate=48000):
    """Reduces noise in the input audio by the given noise file and then gets
    the RMS values of all channels of the input audio.

    @param input_audio: The input audio file to be analyzed.
    @param noise_file: The noise file used to reduce noise in the input audio.
    @param channels: The number of channels in the input audio.
    @param bits: The number of bits of each audio sample.
    @param rate: The sampling rate.
    """
    with tempfile.NamedTemporaryFile() as reduced_file:
        p1 = cmd_utils.popen(
                sox_utils.noise_profile_cmd(
                        noise_file, '-', channels=channels, bits=bits,
                        rate=rate),
                stdout=subprocess.PIPE)
        p2 = cmd_utils.popen(
                sox_utils.noise_reduce_cmd(
                        input_audio, reduced_file.name, '-',
                        channels=channels, bits=bits, rate=rate),
                stdin=p1.stdout)
        cmd_utils.wait_and_check_returncode(p1, p2)
        return get_rms(reduced_file.name, channels, bits, rate)


def cras_rms_test_setup():
    """Setups for the cras_rms_tests.

    To make sure the line_out-to-mic_in path is all green.
    """
    # TODO(owenlin): Now, the nodes are choosed by chrome.
    #                We should do it here.
    cras_utils.set_system_volume(_DEFAULT_PLAYBACK_VOLUME)
    cras_utils.set_selected_output_node_volume(_DEFAULT_PLAYBACK_VOLUME)

    cras_utils.set_capture_gain(_DEFAULT_CAPTURE_GAIN)

    cras_utils.set_system_mute(False)
    cras_utils.set_capture_mute(False)


def generate_rms_postmortem():
    """Generates postmortem for rms tests."""
    try:
        logging.info('audio postmortem report')
        log_loopback_dongle_status()
        logging.info(get_audio_diagnostics())
    except Exception:
        logging.exception('Error while generating postmortem report')


def get_audio_diagnostics():
    """Gets audio diagnostic results.

    @returns: a string containing diagnostic results.

    """
    return cmd_utils.execute([_AUDIO_DIAGNOSTICS_PATH], stdout=subprocess.PIPE)


def get_max_cross_correlation(signal_a, signal_b):
    """Gets max cross-correlation and best time delay of two signals.

    Computes cross-correlation function between two
    signals and gets the maximum value and time delay.
    The steps includes:
      1. Compute cross-correlation function of X and Y and get Cxy.
         The correlation function Cxy is an array where Cxy[k] is the
         cross product of X and Y when Y is delayed by k.
         Refer to manual of numpy.correlate for detail of correlation.
      2. Find the maximum value C_max and index C_index in Cxy.
      3. Compute L2 norm of X and Y to get norm(X) and norm(Y).
      4. Divide C_max by norm(X)*norm(Y) to get max cross-correlation.

    Max cross-correlation indicates the similarity of X and Y. The value
    is 1 if X equals Y multiplied by a positive scalar.
    The value is -1 if X equals Y multiplied by a negative scaler.
    Any constant level shift will be regarded as distortion and will make
    max cross-correlation value deviated from 1.
    C_index is the best time delay of Y that make Y looks similar to X.
    Refer to http://en.wikipedia.org/wiki/Cross-correlation.

    @param signal_a: A list of numbers which contains the first signal.
    @param signal_b: A list of numbers which contains the second signal.

    @raises: ValueError if any number in signal_a or signal_b is not a float.
             ValueError if norm of any array is less than _MINIMUM_NORM.

    @returns: A tuple (correlation index, best delay). If there are more than
              one best delay, just return the first one.
    """
    def check_list_contains_float(numbers):
        """Checks the elements in a list are all float.

        @param numbers: A list of numbers.

        @raises: ValueError if there is any element which is not a float
                 in the list.
        """
        if any(not isinstance(x, float) for x in numbers):
            raise ValueError('List contains number which is not a float')

    check_list_contains_float(signal_a)
    check_list_contains_float(signal_b)

    norm_a = numpy.linalg.norm(signal_a)
    norm_b = numpy.linalg.norm(signal_b)
    logging.debug('norm_a: %f', norm_a)
    logging.debug('norm_b: %f', norm_b)
    if norm_a <= _MINIMUM_NORM or norm_b <= _MINIMUM_NORM:
        raise ValueError('No meaningful data as norm is too small.')

    correlation = numpy.correlate(signal_a, signal_b, 'full')
    max_correlation = max(correlation)
    best_delays = [i for i, j in enumerate(correlation) if j == max_correlation]
    if len(best_delays) > 1:
        logging.warning('There are more than one best delay: %r', best_delays)
    return max_correlation / (norm_a * norm_b), best_delays[0]


def trim_data(data, threshold=0):
    """Trims a data by removing value that is too small in head and tail.

    Removes elements in head and tail whose absolute value is smaller than
    or equal to threshold.
    E.g. trim_data([0.0, 0.1, 0.2, 0.3, 0.2, 0.1, 0.0], 0.2) =
    ([0.2, 0.3, 0.2], 2)

    @param data: A list of numbers.
    @param threshold: The threshold to compare against.

    @returns: A tuple (trimmed_data, end_trimmed_length), where
              end_trimmed_length is the length of original data being trimmed
              from the end.
              Returns ([], None) if there is no valid data.
    """
    indice_valid = [
            i for i, j in enumerate(data) if abs(j) > threshold]
    if not indice_valid:
        logging.warning(
                'There is no element with absolute value greater '
                'than threshold %f', threshold)
        return [], None
    logging.debug('Start and end of indice_valid: %d, %d',
                  indice_valid[0], indice_valid[-1])
    end_trimmed_length = len(data) - indice_valid[-1] - 1
    logging.debug('Trimmed length in the end: %d', end_trimmed_length)
    return (data[indice_valid[0] : indice_valid[-1] + 1], end_trimmed_length)


def get_one_channel_correlation(test_data, golden_data):
    """Gets max cross-correlation of test_data and golden_data.

    Trims test data and compute the max cross-correlation against golden_data.
    Signal can be trimmed because those zero values in the head and tail of
    a signal will not affect correlation computation.

    @param test_data: A list containing the data to compare against golden data.
    @param golden_data: A list containing the golden data.

    @returns: A tuple (max cross-correlation, best_delay) if data is valid.
              Otherwise returns (None, None). Refer to docstring of
              get_max_cross_correlation.
    """
    trimmed_test_data, end_trimmed_length = trim_data(test_data)

    def to_float(samples):
      """Casts elements in the list to float.

      @param samples: A list of numbers.

      @returns: A list of original numbers casted to float.
      """
      samples_float = [float(x) for x in samples]
      return samples_float

    max_cross_correlation, best_delay =  get_max_cross_correlation(
            to_float(golden_data),
            to_float(trimmed_test_data))

    # The reason to add back the trimmed length in the end.
    # E.g.:
    # golden data:
    #
    # |-----------vvvv----------------|  vvvv is the signal of interest.
    #       a                 b
    #
    # test data:
    #
    # |---x----vvvv--------x----------------|  x is the place to trim.
    #   c   d         e            f
    #
    # trimmed test data:
    #
    # |----vvvv--------|
    #   d         e
    #
    # The first output of cross correlation computation :
    #
    #                  |-----------vvvv----------------|
    #                       a                 b
    #
    # |----vvvv--------|
    #   d         e
    #
    # The largest output of cross correlation computation happens at
    # delay a + e.
    #
    #                  |-----------vvvv----------------|
    #                       a                 b
    #
    #                         |----vvvv--------|
    #                           d         e
    #
    # Cross correlation starts computing by aligning the last sample
    # of the trimmed test data to the first sample of golden data.
    # The best delay calculated from trimmed test data and golden data
    # cross correlation is e + a. But the real best delay that should be
    # identical on two channel should be e + a + f.
    # So we need to add back the length being trimmed in the end.

    if max_cross_correlation:
        return max_cross_correlation, best_delay + end_trimmed_length
    else:
        return None, None


def compare_one_channel_correlation(test_data, golden_data, parameters):
    """Compares two one-channel data by correlation.

    @param test_data: A list containing the data to compare against golden data.
    @param golden_data: A list containing the golden data.
    @param parameters: A dict containing parameters for method.

    @returns: A dict containing:
              index: The index of similarity where 1 means they are different
                  only by a positive scale.
              best_delay: The best delay of test data in relative to golden
                  data.
              equal: A bool containing comparing result.
    """
    if 'correlation_threshold' in parameters:
        threshold = parameters['correlation_threshold']
    else:
        threshold = _CORRELATION_INDEX_THRESHOLD

    result_dict = dict()
    max_cross_correlation, best_delay = get_one_channel_correlation(
            test_data, golden_data)
    result_dict['index'] = max_cross_correlation
    result_dict['best_delay'] = best_delay
    result_dict['equal'] = True if (
        max_cross_correlation and
        max_cross_correlation > threshold) else False
    logging.debug('result_dict: %r', result_dict)
    return result_dict


def compare_data_correlation(golden_data_binary, golden_data_format,
                             test_data_binary, test_data_format,
                             channel_map, parameters=None):
    """Compares two raw data using correlation.

    @param golden_data_binary: The binary containing golden data.
    @param golden_data_format: The data format of golden data.
    @param test_data_binary: The binary containing test data.
    @param test_data_format: The data format of test data.
    @param channel_map: A list containing channel mapping.
                        E.g. [1, 0, None, None, None, None, None, None] means
                        channel 0 of test data should map to channel 1 of
                        golden data. Channel 1 of test data should map to
                        channel 0 of golden data. Channel 2 to 7 of test data
                        should be skipped.
    @param parameters: A dict containing parameters for method, if needed.

    @raises: NotImplementedError if file type is not raw.
             NotImplementedError if sampling rates of two data are not the same.
             error.TestFail if golden data and test data are not equal.
    """
    if parameters is None:
        parameters = dict()

    if (golden_data_format['file_type'] != 'raw' or
        test_data_format['file_type'] != 'raw'):
        raise NotImplementedError('Only support raw data in compare_data.')
    if (golden_data_format['rate'] != test_data_format['rate']):
        raise NotImplementedError(
                'Only support comparing data with the same sampling rate')
    golden_data = audio_data.AudioRawData(
            binary=golden_data_binary,
            channel=golden_data_format['channel'],
            sample_format=golden_data_format['sample_format'])
    test_data = audio_data.AudioRawData(
            binary=test_data_binary,
            channel=test_data_format['channel'],
            sample_format=test_data_format['sample_format'])
    compare_results = []
    for test_channel, golden_channel in enumerate(channel_map):
        if golden_channel is None:
            logging.info('Skipped channel %d', test_channel)
            continue
        test_data_one_channel = test_data.channel_data[test_channel]
        golden_data_one_channel = golden_data.channel_data[golden_channel]
        result_dict = dict(test_channel=test_channel,
                           golden_channel=golden_channel)
        result_dict.update(
                compare_one_channel_correlation(
                        test_data_one_channel, golden_data_one_channel,
                        parameters))
        compare_results.append(result_dict)
    logging.info('compare_results: %r', compare_results)
    for result in compare_results:
        if not result['equal']:
            error_msg = ('Failed on test channel %d and golden channel %d with '
                         'index %f') % (
                                 result['test_channel'],
                                 result['golden_channel'],
                                 result['index'])
            logging.error(error_msg)
            raise error.TestFail(error_msg)
    # Also checks best delay are exactly the same.
    best_delays = set([result['best_delay'] for result in compare_results])
    if len(best_delays) > 1:
        error_msg = 'There are more than one best delay: %s' % best_delays
        logging.error(error_msg)
        raise error.TestFail(error_msg)


class _base_rms_test(test.test):
    """Base class for all rms_test """

    def postprocess(self):
        super(_base_rms_test, self).postprocess()

        # Sum up the number of failed constraints in each iteration
        if sum(len(x) for x in self.failed_constraints):
            generate_rms_postmortem()


class chrome_rms_test(_base_rms_test):
    """Base test class for audio RMS test with Chrome.

    The chrome instance can be accessed by self.chrome.
    """
    def warmup(self):
        super(chrome_rms_test, self).warmup()

        # Not all client of this file using telemetry.
        # Just do the import here for those who really need it.
        from autotest_lib.client.common_lib.cros import chrome

        self.chrome = chrome.Chrome(init_network_controller=True)

        # The audio configuration could be changed when we
        # restart chrome.
        try:
            cras_rms_test_setup()
        except Exception:
            self.chrome.browser.Close()
            raise


    def cleanup(self, *args):
        try:
            self.chrome.close()
        finally:
            super(chrome_rms_test, self).cleanup()

class cras_rms_test(_base_rms_test):
    """Base test class for CRAS audio RMS test."""

    def warmup(self):
        super(cras_rms_test, self).warmup()
        cras_rms_test_setup()


class alsa_rms_test(_base_rms_test):
    """Base test class for ALSA audio RMS test.

    Note the warmup will take 10 seconds and the device cannot be used before it
    returns.
    """
    def warmup(self):
        super(alsa_rms_test, self).warmup()

        cras_rms_test_setup()
        # We need CRAS to initialize the volume and gain.
        cras_utils.playback(playback_file="/dev/zero", duration=1)
        # CRAS will release the device after 10 seconds.
        time.sleep(11)