Disclosed are a signal compression apparatus and method for dynamic compressive sensing, including: a signal input unit configured to simultaneously output an input signal, and generate and output a linear measurement reference signal based on the input signal; a linear transform unit configured to receive the linear measurement reference signal and variably generate a linear measurement matrix value according to the linear measurement reference signal; a signal compressor configured to output a compressed signal for the input signal based on the generated linear measurement matrix value; and a signal processor configured to reconstruct the compressed signal and perform spectrum sensing of the input signal. Accordingly, it is possible to increase signal compression efficiency of compressive sensing and improve signal reconstruction performance by using sparsity information of the input signal, and also by performing signal compression by using a compression matrix suitable for the sparsity information of the input signal.