We introduce SynSE (Syntactically guided Skeleton Embeddings), a novel syntactically guided generative approach for Zero-Shot Learning (ZSL). Our end-to-end approach learns progressively refined generative embedding spaces constrained within and across the involved modalities (visual, language). The inter-modal constraints are defined between action sequence embedding and embeddings of Parts of Speech (PoS) tagged words in the corresponding action description. We deploy SynSE for the task of skeleton-based action sequence recognition. Our design choices enable SynSE to generalize compositionally, i.e., recognize sequences whose action descriptions contain words not encountered during training. We also extend our approach to the more challenging Generalized Zero-Shot Learning (GZSL) problem via a confidence-based gating mechanism. We are the first to present zero-shot skeleton action recognition results on the large-scale NTU-60 and NTU-120 skeleton action datasets with multiple splits. Our results demonstrate SynSE’s state-of-the-art performance in both ZSL and GZSL settings compared to strong baselines on the NTU-60 and NTU-120 datasets.

We analyze the error performance of minimum selection generalized selection combining (MS-GSC), in which the minimum number of diversity branches are selected such that their combined signal-to-noise ratio (SNR) is above a given threshold. A flat Rayleigh fading channel with independent and distinctly distributed branch SNR is considered. By transforming the ordered instantaneous branch SNR to their differences, we derive the distribution of the number of selected branches in closed-form. We then modify the derivation of this distribution to get the characteristic function (c.f.) of the combiner output SNR. This c.f. is used to obtain the symbol error probability for different coherent digital modulation schemes.

We analyze the error performance of an improved hybrid selection/maximal-ratio combining (H-S/MRC) technique called minimum selection H-S/MRC in flat Rayleigh fading for coherent digital modulation schemes. Here the minimum number of diversity branches are selected such that their combined signal-to-noise ratio is above a given threshold. We derive a closed-form expression for the distribution of the number of selected branches. This distribution is used to obtain the symbol error probability.

This paper analyzes the error performance of an improved generalized selection combining (GSC) technique called minimum selection GSC (MS-GSC) in flat Rayleigh fading for coherent digital modulation schemes. Here the minimum number of diversity branches is selected such that their combined signal-to-noise ratio (SNR) is above a given threshold. The threshold chosen here is determined by the combiner output SNR rather than individual branch SNRs as done in absolute threshold GSC (AT-GSC) and normalized threshold GSC (NT-GSC). As a result, MS-GSC uses less resource than AT-GSC or NT-GSC to give the same performance. We obtain a closed-form expression for the distribution of the number of selected branches followed by a closed-form expression for the characteristic function (c.f.) of the combiner output SNR. This c.f. is used to obtain the symbol error probability for different digital modulation schemes. Numerical results for M-ary phase-shift keying are presented as an example.