Speech Synthesis From Neural Decoding Of Spoken Sentences