Music has become one of daily essentials in our modern life. Most people enjoy listening to their favorite music and spend considerable time with music. But, one step forward to playing music and composing music are still not easy for everyone. An automatic music composition system ("Orpheus") for given lyrics will be introduced as a combination of music theory, linguistics and a speech recognition algorithm. Also, an automatic performance rendering system ("Polyhymnia") of piano pieces will be discussed from a probabilistic approach using conditional random field. Automatic orchestral accompaniment ("Euridice") may be useful for those who plays the piano but cannot have a chance to play with orchestras as another application of HMM (Hidden Markov Model) from the speech recognition technology. If time allows, other topics related to music signal processing including harmonic-percussive sound separation, sound separation of polyphonic audio signals, music genre recognition, chord recognition, etc. will also be discussed.