public override WAVSound GenerateWord(string word) { int wordIndex = wordToSoundMappingList.FindIndex(m => m.Word == word); WAVSound wordSound = null; if (wordIndex >= 0) { WordToSoundMapping wordToSoundMapping = wordToSoundMappingList[wordIndex]; List <WAVSound> wavSoundList = new List <WAVSound>(); foreach (string soundName in wordToSoundMapping.SoundNameList) { FormantSpecification formantSpecification = SpecificationList.Find(s => s.Name == soundName); if (formantSpecification != null) { formantSpecification.GenerateSettingsSequence(); WAVSound sound = GenerateSound(formantSpecification); wavSoundList.Add(sound); } } if (wavSoundList.Count > 0) { wordSound = WAVSound.Join(wavSoundList, null); } } return(wordSound); }
public FormantSpecification Copy() { FormantSpecification copiedFormantSpecification = new FormantSpecification(this.fundamentalFrequency, this.samplingFrequency); foreach (FormantSettings formantSettings in this.formantSettingsList) { FormantSettings copiedFormantSettings = formantSettings.Copy(); copiedFormantSpecification.FormantSettingsList.Add(copiedFormantSettings); } return(copiedFormantSpecification); }
public WAVSound GenerateSound(FormantSpecification formantSpecification) { double duration = formantSpecification.GetDuration(); int samplingFrequency = formantSpecification.SamplingFrequency; double sampleTime = 1.0 / (double)samplingFrequency; double nominalPulsePeriod = 1 / (double)formantSpecification.FundamentalFrequency; int numberOfSamples = (int)Math.Round(duration / sampleTime); // Generate voiced pulse train: This requires running through the sequence // of formantSettings in the formantSpecification, in order to determine // the pitch (and its inverse, the pulse period) at each time: List <int> voicedPulseSpacingList = formantSpecification.GeneratePulseSpacingList(); List <double> voicedPulseTrain = new List <double>(); for (int ii = 0; ii < numberOfSamples; ii++) { voicedPulseTrain.Add(0); } // To be adjusted below. int sampleIndex = 0; while (sampleIndex < voicedPulseTrain.Count) { voicedPulseTrain[sampleIndex] = 1.0; sampleIndex += voicedPulseSpacingList[sampleIndex]; } // Generate unvoiced pulse train: if (randomNumberGenerator == null) { randomNumberGenerator = new Random(); } // if (gaussian == null) { gaussian = new GaussianDistribution(0, 0.05, -1); } List <double> unvoicedPulseTrain = new List <double>(); for (int ii = 0; ii < numberOfSamples; ii++) { unvoicedPulseTrain.Add(0); if (randomNumberGenerator.NextDouble() < 0.5) { unvoicedPulseTrain[ii] = -WHITE_NOISE_LEVEL + 2 * WHITE_NOISE_LEVEL * randomNumberGenerator.NextDouble(); // gaussian.GetSample(); } } // Set up sinusoids: int numberOfSinusoids = formantSpecification.GetNumberOfSinusoids(); List <DampedSinusoid> sinusoidList = new List <DampedSinusoid>(); for (int iSinusoid = 0; iSinusoid < numberOfSinusoids; iSinusoid++) { DampedSinusoid sinusoid = new DampedSinusoid(samplingFrequency); sinusoidList.Add(sinusoid); } // Prepare for storing pitch: if (storePitch) { // pitchList = new List<double>(); timePitchPeriodList = new List <List <double> >(); } // Generate the relative amplitude list: Must be done separately, to handle // transitions: List <double> relativeAmplitudeList = formantSpecification.GenerateRelativeAmplitudeList(); // Generate the unscaled samples: List <double> unscaledSampleList = new List <double>(); double time = 0; sampleIndex = 0; while (sampleIndex < numberOfSamples) { time = sampleIndex * sampleTime; FormantSettings formantSettings = formantSpecification.GetInterpolatedSettings(sampleIndex); for (int iSinusoid = 0; iSinusoid < sinusoidList.Count; iSinusoid++) { sinusoidList[iSinusoid].SetParameters(formantSettings.AmplitudeList[iSinusoid], formantSettings.FrequencyList[iSinusoid], formantSettings.BandwidthList[iSinusoid]); } double x = formantSettings.VoicedFraction * voicedPulseTrain[sampleIndex] + (1 - formantSettings.VoicedFraction) * unvoicedPulseTrain[sampleIndex]; // 20170407 if (storePitch) { if (formantSettings.VoicedFraction > minimumVoicedFractionForPitch) { if (voicedPulseTrain[sampleIndex] != 0) // Define the pitch only at pulse spikes { double pitch = samplingFrequency / voicedPulseSpacingList[sampleIndex]; // pitchList.Add(pitch); double pitchPeriod = voicedPulseSpacingList[sampleIndex] * sampleTime; timePitchPeriodList.Add(new List <double>() { time, pitchPeriod }); } /* else { pitchList.Add(-1); } * } * else * { * pitchList.Add(-1); // < 0 => pitch not defined */ } } double deltaTime = formantSpecification.DeltaTimeList[sampleIndex]; double relativeAmplitude = relativeAmplitudeList[sampleIndex]; // formantSettings.GetRelativeAmplitude(deltaTime); double sample = 0; for (int iSinusoid = 0; iSinusoid < sinusoidList.Count; iSinusoid++) { sample += sinusoidList[iSinusoid].Next(x); } sample *= relativeAmplitude * volume; unscaledSampleList.Add(sample); sampleIndex++; } // Next generate the scaled samples List <Int16> sampleList = new List <Int16>(); for (int ii = 0; ii < numberOfSamples; ii++) { if (unscaledSampleList[ii] > 1) { unscaledSampleList[ii] = 1; } else if (unscaledSampleList[ii] < -1) { unscaledSampleList[ii] = -1; } Int16 sample = (Int16)Math.Round(32767 * unscaledSampleList[ii]); // Some ugly hard-coding here... sampleList.Add(sample); } List <List <Int16> > twoChannelSampleList = new List <List <Int16> >(); twoChannelSampleList.Add(sampleList); WAVSound wavSound = new WAVSound("Test", samplingFrequency, 1, 16); // Some ugly hard-coding here... wavSound.GenerateFromSamples(twoChannelSampleList); return(wavSound); }