Abstract: Bipolar junction transistor (BJT) structures are provided. First and second well regions are formed over a dielectric layer. A plurality of first and second gate-all-around (GAA) field-effect transistors are formed over a first well region. A plurality of third GAA field-effect transistors are formed over the second well region. Source/drain features of the first and third GAA field-effect transistors and the second well region have a first conductivity type. Source/drain features of the second GAA field-effect transistors and the first well region have a second conductivity type that is different from the first conductivity type. A first PN junction of a first BJT device is formed between the source/drain features of the first GAA field-effect transistors and the first well region, and a second PN junction of the first BJT device is formed between the first well region and the second well region.

Abstract: The invention provides a method for automatically identifying a modulation mode for a digital communication signal. The method comprises the following steps of: S1: preprocessing a training digital modulation signal; S2: extracting a characteristic value of the training digital modulation signal; S3: constructing a strong classifier by Bagging learning; and S4: preprocessing a modulation signal to be tested and extracting a corresponding characteristic parameter, and then inputting the characteristic parameter into the strong classifier in the S3 to classify and identify the modulation signal. According to the method, by processing all characteristic values in parallel, a success rate of identification has little to do with an identification performance of a single characteristic parameter and is mainly related to an overall performance, so that the identification to the modulation mode has the advantages of fast identification speed and high success rate of identification.

Abstract: The invention provides a method for automatically identifying a modulation mode for a digital communication signal. The method comprises the following steps of: S1: preprocessing a training digital modulation signal; S2: extracting a characteristic value of the training digital modulation signal; S3: constructing a strong classifier by Bagging learning; and S4: preprocessing a modulation signal to be tested and extracting a corresponding characteristic parameter, and then inputting the characteristic parameter into the strong classifier in the S3 to classify and identify the modulation signal. According to the method, by processing all characteristic values in parallel, a success rate of identification has little to do with an identification performance of a single characteristic parameter and is mainly related to an overall performance, so that the identification to the modulation mode has the advantages of fast identification speed and high success rate of identification.