TR2 and TR3 are biased in class B, which means that they are normally non conducting when there is no signal input. TR1 and T1 form a phase splitter, opposite polarity signals appearing at the ends of T1 secondary.
When there is a signal in ,TR2 conducts on the positive half of the waveform only and TR3 conducts on the negative half. Current flows down through the top half of T2 primary when TR2 conducts, and up through the bottom half when TR3 conducts. Both halves are combined in the secondary of T2 to produce an amplified version of the input.
R1 and R2 apply a small forward bias to avoid crossover distortion, which is shown in the lower waveform. This distortion is caused by the non-linear characteristics of the emitter/base junction. (for more details see the junction diode Vb/Ib curves).
The advantage of this circuit is that there is very little current taken if
there is no signal input.