The present technique relates to an earhole-wearable sound collection device, a signal processing device, and a method for realizing sound collection at a high S/N ratio. In the earhole-wearable sound collection device, a microphone that collects emitted speech voice is provided in a space that is substantially sealed off from outside and connects to an ear canal of the wearer (the speaker). With the microphone being located in the space sealed off from outside, emitted speech voice that propagates through the ear canal of the wearer is collected. In a sound collection signal obtained through the ear canal, the emitted speech voice component is dominant over the noise component particularly at low frequencies. Therefore, the S/N ratio of an emitted speech voice collection signal is improved by an equalizing process for reducing muffled sound that is generated when sound is collected through the ear canal on the sound collection signal.