Hydrocarbon gas sensing is a challenging task using laser absorption spectroscopy due to the complex and broad structure of absorption lines. This application requires quick, accurate and highly sensitive detection of hydrocarbon gases concentrations. In this paper, a compact photoacoustic spectrophone was developed to simultaneously measure methane, propane and isobutane. This spectrophone uses wavelength modulation spectroscopy (WMS) with a single acoustic resonator and a single DFB laser emitting at 3368 nm, which greatly reduces the system complexity without using time-division multiplexing technology for multi-gas sensing. Due to the complex and broadband absorption of hydrocarbon gases, a novel signal processing method based on multilinear regression with Ridge regression (MLR-RG) is proposed to reduce the measurement error caused by the nonlinearity of spectra signal. For single gas measurement, the detection limits of methane, propane, and isobutane are determined to be 828 ppb, 419 ppb, and 619 ppb (SNR = 1, integration time = 20 s), respectively. For simultaneous multi-gas sensing in a gaseous mixture, the detection limits of propane and isobutane are determined to be 7 ppb, 68 ppb with an integration time of 860 s, 460 s, respectively. The measurement accuracy of propane and isobutane using MLR-RG is higher than that of ordinary least squares regression and partial least squares regression by 75% and 60%, respectively. The proposed algorithm based on MLR-RG provides a promising approach to process the broad overlapping absorption spectra for accurately retrieving hydrocarbon gases concentrations.