Automatic speech recognition (ASR) systems often need to identify the speaker's gender when processing speech data, which is usually part of speech feature analysis. For example, in ASR annotation, the speaker separation and annotation stage assigns attribute labels such as gender to different speakers in the audio to help the model learn the contextual information of the speech. This annotation helps improve the accuracy of speech interaction, such as in multi-person dialogues or conference transcription scenarios. From a technical perspective, certain physical characteristics of speech signals (such as pitch and formants) are statistically correlated with gender; therefore, ASR systems can use these features to infer gender.