Baum-Welch Learning in Hidden Markov Model using Multivariate Gaussian Model

[James Kenway]

I've used this for common normal distribution for continuous data:

// Specify a initial normal distribution for the samples.
            NormalDistribution density = new NormalDistribution();

            // Creates a continuous hidden Markov Model with two states organized in a forward
            //  topology and an underlying univariate Normal distribution as probability density.
            var model = new HiddenMarkovModel<NormalDistribution>(new Ergodic(2), density);

            // Configure the learning algorithms to train the sequence classifier until the
            // difference in the average log-likelihood changes only by as little as 0.0001
            var teacher = new BaumWelchLearning<NormalDistribution>(model)
            {
                Tolerance = 0.001,
                Iterations = 0,
            };

            // Fit the model
            double likelihood = teacher.Run(data);

            // See the log-probability of the sequences learned
            double a1 = model.Evaluate(new[] { 0.999999999999928, 0 , 0.999999999999988 , 0 , 0.999999999999988 }); // -0.12799388666109757
            return a1;

How to use Multivariate Gaussian/normal distribution for continuous data? Thank you.