Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Are the SpeakProgressEventArgs of the SpeechSynthesizer inaccurate?

Using the System.Speech.Synthesis.SpeechSynthesizer class in .Net 3.5, the AudioPosition property of the SpeakProgressEventArgs appears to be inaccurate.

The following code produces the following output:

Code:

using System;
using System.Speech.Synthesis;
using System.Threading;

namespace SpeechTest
{
    class Program
    {
        static ManualResetEvent speechDoneEvent = new ManualResetEvent(false);

        static void Main(string[] args)
        {
            SpeechSynthesizer synthesizer = new SpeechSynthesizer();

            synthesizer.SpeakProgress += new EventHandler<SpeakProgressEventArgs>(synthesizer_SpeakProgress);

            synthesizer.SpeakCompleted += new EventHandler<SpeakCompletedEventArgs>(synthesizer_SpeakCompleted);

            synthesizer.SetOutputToWaveFile("Test.wav");

            synthesizer.SpeakAsync("This holiday season, support the music you love by shopping at Made in Washington, online and at one of five local stores. Made in Washington chocolates, bountiful gift baskets and ornaments are the perfect holiday gifts for family, friends and co-workers.");

            speechDoneEvent.WaitOne();
        }

        static void synthesizer_SpeakCompleted(object sender, SpeakCompletedEventArgs e)
        {
            speechDoneEvent.Set();
        }

        static void synthesizer_SpeakProgress(object sender, SpeakProgressEventArgs e)
        {
            Console.WriteLine("SpeakProgress: AudioPosition=" + e.AudioPosition + ",\tCharacterPosition=" + e.CharacterPosition + ",\tCharacterCount=" + e.CharacterCount + ",\tText=" + e.Text);
        }
    }
}

Output:

SpeakProgress: AudioPosition=00:00:00.0043750,  CharacterPosition=0,    CharacterCount=4,       Text=This
SpeakProgress: AudioPosition=00:00:00.2925625,  CharacterPosition=5,    CharacterCount=7,       Text=holiday
SpeakProgress: AudioPosition=00:00:00.9086250,  CharacterPosition=13,   CharacterCount=6,       Text=season
SpeakProgress: AudioPosition=00:00:01.9421250,  CharacterPosition=21,   CharacterCount=7,       Text=support
SpeakProgress: AudioPosition=00:00:02.5621250,  CharacterPosition=29,   CharacterCount=3,       Text=the
SpeakProgress: AudioPosition=00:00:02.6760625,  CharacterPosition=33,   CharacterCount=5,       Text=music
SpeakProgress: AudioPosition=00:00:03.2648125,  CharacterPosition=39,   CharacterCount=3,       Text=you
SpeakProgress: AudioPosition=00:00:03.5199375,  CharacterPosition=43,   CharacterCount=4,       Text=love
SpeakProgress: AudioPosition=00:00:03.8435625,  CharacterPosition=48,   CharacterCount=2,       Text=by
SpeakProgress: AudioPosition=00:00:04.0701875,  CharacterPosition=51,   CharacterCount=8,       Text=shopping
SpeakProgress: AudioPosition=00:00:04.6840625,  CharacterPosition=60,   CharacterCount=2,       Text=at
SpeakProgress: AudioPosition=00:00:04.8036250,  CharacterPosition=63,   CharacterCount=4,       Text=Made
SpeakProgress: AudioPosition=00:00:05.0698125,  CharacterPosition=68,   CharacterCount=2,       Text=in
SpeakProgress: AudioPosition=00:00:05.2521250,  CharacterPosition=71,   CharacterCount=10,      Text=Washington
SpeakProgress: AudioPosition=00:00:06.2961875,  CharacterPosition=83,   CharacterCount=6,       Text=online
SpeakProgress: AudioPosition=00:00:07.0540625,  CharacterPosition=90,   CharacterCount=3,       Text=and
SpeakProgress: AudioPosition=00:00:07.3331250,  CharacterPosition=94,   CharacterCount=2,       Text=at
SpeakProgress: AudioPosition=00:00:07.6818750,  CharacterPosition=97,   CharacterCount=3,       Text=one
SpeakProgress: AudioPosition=00:00:08.0598750,  CharacterPosition=101,  CharacterCount=2,       Text=of
SpeakProgress: AudioPosition=00:00:08.2163750,  CharacterPosition=104,  CharacterCount=4,       Text=five
SpeakProgress: AudioPosition=00:00:08.5971875,  CharacterPosition=109,  CharacterCount=5,       Text=local
SpeakProgress: AudioPosition=00:00:09.0243750,  CharacterPosition=115,  CharacterCount=6,       Text=stores
SpeakProgress: AudioPosition=00:00:10.5325625,  CharacterPosition=123,  CharacterCount=4,       Text=Made
SpeakProgress: AudioPosition=00:00:10.7700625,  CharacterPosition=128,  CharacterCount=2,       Text=in
SpeakProgress: AudioPosition=00:00:10.9377500,  CharacterPosition=131,  CharacterCount=10,      Text=Washington
SpeakProgress: AudioPosition=00:00:11.6708125,  CharacterPosition=142,  CharacterCount=10,      Text=chocolates
SpeakProgress: AudioPosition=00:00:12.9798750,  CharacterPosition=154,  CharacterCount=9,       Text=bountiful
SpeakProgress: AudioPosition=00:00:13.6303125,  CharacterPosition=164,  CharacterCount=4,       Text=gift
SpeakProgress: AudioPosition=00:00:14.0959375,  CharacterPosition=169,  CharacterCount=7,       Text=baskets
SpeakProgress: AudioPosition=00:00:14.7848125,  CharacterPosition=177,  CharacterCount=3,       Text=and
SpeakProgress: AudioPosition=00:00:15.0507500,  CharacterPosition=181,  CharacterCount=9,       Text=ornaments
SpeakProgress: AudioPosition=00:00:15.7195000,  CharacterPosition=191,  CharacterCount=3,       Text=are
SpeakProgress: AudioPosition=00:00:15.9872500,  CharacterPosition=195,  CharacterCount=3,       Text=the
SpeakProgress: AudioPosition=00:00:16.1488750,  CharacterPosition=199,  CharacterCount=7,       Text=perfect
SpeakProgress: AudioPosition=00:00:16.7275000,  CharacterPosition=207,  CharacterCount=7,       Text=holiday
SpeakProgress: AudioPosition=00:00:17.3336875,  CharacterPosition=215,  CharacterCount=5,       Text=gifts
SpeakProgress: AudioPosition=00:00:17.9813125,  CharacterPosition=221,  CharacterCount=3,       Text=for
SpeakProgress: AudioPosition=00:00:18.2216875,  CharacterPosition=225,  CharacterCount=6,       Text=family
SpeakProgress: AudioPosition=00:00:19.0973750,  CharacterPosition=233,  CharacterCount=7,       Text=friends
SpeakProgress: AudioPosition=00:00:19.7726250,  CharacterPosition=241,  CharacterCount=3,       Text=and
SpeakProgress: AudioPosition=00:00:19.9655625,  CharacterPosition=245,  CharacterCount=10,      Text=co-workers
SpeakProgress: AudioPosition=00:00:20.2518750,  CharacterPosition=245,  CharacterCount=10,      Text=co-workers

However, the duration of the .wav file produced is 15.69 seconds. The same behavior occurs if you output to a Stream or to null.

The documentation for the property says the property is "A TimeSpan object that represents the time position of the event in the audio output stream".

Should it be an accurate time indicating the time the word is started or finished speaking in the output file, or am I misinterpreting it?

like image 417
jameswelle Avatar asked Nov 12 '09 00:11

jameswelle


1 Answers

the audioPosition depends on the selected voice of the speech synthesizer. For some Microsoft voices, such as Anna, Zira, David, Hazel, as I have experienced, the supported audio format is a 16000Hz PCM. So the following solution can correct the auido position:

var format = 
new System.Speech.AudioFormat.SpeechAudioFormatInfo(EncodingFormat.Pcm, 
                                                    16000, 16, 1, 32000, 2, null);
synthesizer.SetOutputToWaveFile("Test.wav", format);

if you note, the default sample rate of the SetOutputToWaveFile is 22050, and the ratio of the correct time (15.69) to the time shown by AudipPosition (20.25) is about 0.77. If you multiply this ratio by 22050 you get about 16000, which is the correct sample rate.

like image 168
Ahmad Avatar answered Sep 30 '22 21:09

Ahmad